how public information affects asymmetrically informed lenders
TRANSCRIPT
Board of Governors of the Federal Reserve System
International Finance Discussion Papers
Number 1125
November 2014
How Public Information Affects Asymmetrically Informed Lenders:
Evidence from a Credit Registry Reform
M. Ali Choudhary and Anil K. Jain
NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate
discussion and critical comment. References to International Finance Discussion Papers (other than an
acknowledgment that the writer has had access to unpublished material) should be cleared with the
author or authors. Recent IFDPs are available on the Web at www.federalreserve.gov/pubs/ifdp/. This
paper can be downloaded without charge from Social Science Research Network electronic library at
www.ssrn.com.
HOW PUBLIC INFORMATION AFFECTS ASYMMETRICALLYINFORMED LENDERS: EVIDENCE FROM A CREDIT REGISTRY
REFORM
M. ALI CHOUDHARY∗ AND ANIL K. JAIN†
Abstract. We exploit exogenous variation in the amount of public information avail-able to banks about a firm to empirically evaluate the importance of adverse selectionin the credit market. A 2006 reform introduced by the State Bank of Pakistan (SBP)reduced the amount of public information available to Pakistani banks about a firm’screditworthiness. Prior to 2006, the SBP published credit information not only about thefirm in question but also (aggregate) credit information about the firm’s group (wherethe group was defined as the set of all firms that shared one or more director with the firmin question). After the reform, the SBP stopped providing the aggregate group-level in-formation. We propose a model with differentially informed banks and adverse selection,which generates predictions on how this reform is expected to affect a bank’s willingnessto lend. The model predicts that adverse selection leads less informed banks to reducelending compared to more informed banks. We construct a measure for the amount ofinformation each lender has about a firm’s group using the set of firm-bank lending pairsprior to the reform. We empirically show those banks with private information abouta firm lent relatively more to that firm than other, less-informed banks following thereform. Remarkably, this reduction in lending by less informed banks is true even forbanks that had a pre-existing relationship with the firm, suggesting that the strength ofprior relationships does not eliminate the problem of imperfect information.
Keywords: Information, Credit registries, Financial IntermediationJEL codes: G14, 016
Date: November 2014.∗State Bank of Pakistan. Email: [email protected] and Federal Reserve Board ofGovernors†. Email: [email protected]. We wish to thank Abhijit Banerjee, Emily Breza,Rebecca Dizon-Ross, Esther Duflo, Ben Feigenberg, Robert Gibbons, Benjamin Golub, ConradMiller, Ben Olken, Jennifer Peck, Adam Sacarny, Annalisa Scognamiglio, Ashish Shenoy,Tavneet Suri, and Robert Townsend for their amazing support on the paper. We thankbanking-cluster executives of the State Bank of Pakistan for their support and candidfeedback. We also would like to thank commercial bank senior executives for participatingin our interviews. We also very grateful to seminar participants at North-EasternUniversity Development Conference (NEUDC), Fuqua Business School, Olin Business School,Haas College of Business, Bank of England, Federal Reserve Board of Governors, Universityof Illinois at Urbana-Champaign (UIUC) and Simon Fraser University. All mistakes areour own. The findings and conclusions in this paper are solely the responsibility of theauthors and should not be interpreted as reflecting the views of the Board of Governorsof the Federal Reserve System, the views of any other person associated with the FederalReserve System or the views of the State Bank of Pakistan.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 1
1. Introduction
Credit markets are crucial to economic growth but asymmetric information may hinder
their effectiveness. Jaffee and Russell [1976] and Stiglitz and Weiss [1981] theoretically
show that adverse selection is a factor that can substantially constrain the effectiveness of
credit markets, yet the extent to which this factor is a problem in credit markets remains
largely unknown. It is possible, for example, that lenders face other binding constraints
(such as those due to moral hazard) that make adverse selection less relevant.
The key empirical challenge to isolating the importance of adverse selection in the real
world is that the variation in the individual observations is unlikely to be exogenous. For
example, lending could drive the asymmetry of information rather than the reverse: one
bank may know more about a firm than another bank simply because it has a lent to that
firm in the past.
An innovative recent paper by Karlan and Zinman [2009] on adverse selection (and moral
hazard) in credit markets tackles one aspect of this issue. Karlan et. al examine if
exogenously lowering the interest rate changes selection of the borrower pool in ways
that the lender cannot observe. While they are able to highlight borrower behavior,
they are still unable to observe changes in lender behavior in the presence of information
asymmetries. In particular, they are not able to demonstrate that lenders react to these
changing unobservables by altering the terms they offer – since the set of contracts offered
by the lenders is fixed – which is the key mechanism behind credit market failure in the
theoretical literature.
The purpose of this paper is to show that banks do change their lending behavior in reac-
tion to a change in the distribution of information about a particular firm. In particular,
among a group of differentially informed lenders, we ask: What is the effect of reducing
public information about a borrower’s creditworthiness?
A regulatory reform by the State Bank of Pakistan (SBP) in April 2006 offers a unique
opportunity to answer this question. The reform exogenously reduced the amount of
public information available to lenders about a firm’s creditworthiness and did so in a way
that varied across firm-lender pairs. Specifically, the change limited a lender’s capacity to
procure information about a firm’s relationships to other firms.
Until April 2006, the SBP had supplemented credit information about prospective bor-
rowers with information about that firm’s “group.” This group was defined as all other
firms which shared at least one director with the borrowing firm.1 But in April 2006,
1Firms within a group have complex interfirm relationships which subsequently have important economicimplications. There is both theoretical and empirical evidence that interfirm relationships may be amechanism for tax reduction (Desai and Dharmapala [2009]), tunneling (Bertrand et al. [2002]), risk
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 2
the SBP stopped defining a firm’s group and in doing so stopped providing group-specific
information (see section 3.1 for more detail).
Lenders value such information about the credit of other firms in a borrower’s network
because assets and profits may be transferred within a group of firms – especially if one
firm is in financial difficulty. For instance, when Lehman Brothers Holdings Inc. filed for
bankruptcy, an unscheduled transfer of $8 billion occurred from the European operations
(Lehman Brothers International) to the US operations.2 In all, the firm’s bankruptcy led
to $38 billion in claims among the various arms of Lehman Brothers and took over three
years of litigation to settle.3
We take advantage of the natural experiment the SBP reform generated: We use a
difference-in-difference methodology to estimate the causal effect of the reduction in pub-
lic information on a bank’s willingness to lend.4 In particular, we exploit variation in
the impact of the policy across firm-bank pairs generated by each bank’s other lending
relationships. Suppose a firm borrowed from two different banks in Dec 2004. Following
the reform, we predict that one bank – the informed bank – could have more information
about the firm’s actual liabilities, if other members of the firm’s group also borrow from
that bank. We then compare how the loans that the firm receives from the informed and
the uninformed banks changes after the reform. In addition, we examine whether the
reform affected a firm’s ability to access total credit by comparing loan amounts received
by firms who had informed lenders, and those who did not, before and after the policy
reform.
Our main result shows that banks with private information about other firms in a firm’s
group lent more to that firm than other, less-informed lenders did, after the reform, both
on an extensive margin (5.4 percent more likely to renew the firm’s loan) and on an
intensive margin (larger loans). This is the primary evidence that the reduction in public
information amplified the problem of adverse selection in the credit market.
Second, those firms that borrowed from informed lenders were likely to borrow 11-14%
more than those firms who did not have access to informed lenders. In other words, both
the level of credit and its source were affected.
Third, using the distribution of firm-bank pairs and the strength of interfirm relations we
construct a measure for the quality of private information each bank possessed about a
firm’s group prior to the reform. Following the reform, those lenders with greater private
sharing (Khwaja and Mian [2005a]), the efficient working of organizations (Williamson [1975, 1981]) andinternal capital markets (Stein [1997], Almeida and Wolfenzon [2006], Gopalan et al. [2007]).2“Outcry Grows Over Transfer of U.K. Funds by Lehman”, Wall Street Journal, published September 22,2008. http://online.wsj.com/article/SB122204286442761375.html3“Lehman Ends $38 Billion Standoff”’, Wall Street Journal, published October 5, 2012. http://online.
wsj.com/article/SB10000872396390444223104578038234046506220.html4For brevity, in this paper we use “bank” to denote any financial institution.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 3
information about a firm’s group were more likely to renew a firm’s loan.5 This further
supports the claim that a greater information differential between lenders leads to a greater
disparity in the likelihood of a loan being renewed.
Fourth, there is substantial heterogeneity in the measured effect across firms: The smaller
firms in a group were the most disadvantaged from the information change. There was no
effect on the largest firms in a group, which suggests either that information about these
borrowers was already pervasive, or that the relative cost of procuring the information
was lower.
Fifth, those firms which had negative information in their credit reports in December
2004 were the most affected by the change in information reporting. They were 18% more
likely to renew their loan at an informed lender. Those firms with a poor credit history are
likely to be the most risky for a bank, making public information all the more important
for these firms. A firm may have overdue loans, but if the rest of the group is prompt
in their repayment, it may signal a sufficiently creditworthy borrower. However, if both
the firm and the group are overdue, this could signal wider systemic issues in the firm’s
creditworthiness.
The rest of the paper proceeds as follows: Section 2 reviews the relevant literature on
credit markets, and section 3 describes the institutional background, and explains the
SBP’s reasoning in instituting the change in lending policy. Section 4 presents a stylized
model to explain the results and shows how information asymmetries between lenders can
be important. Section 5 outlines our econometric framework for analysing the importance
of information asymmetries across lenders and details our results. Section 6 outlines the
effects of the reform on a firm’s ability to procure credit and section 7 presents a summary
and concluding remarks for future research.
2. Literature Review
This paper’s main contribution is to empirically assess the impact of adverse selection in
credit markets with differentially informed lenders. There is a long theoretical literature
describing how asymmetric information in credit markets causes lenders to alter what
contracts they offer. Compared to an environment with full information, this can lead
to the misallocation of capital (Jaffee and Russell [1976], Stiglitz and Weiss [1981, 1983],
De Meza and Webb [1987]), even to the complete unraveling of the credit market (Akerlof
[1970]). Yet, the empirical evidence for adverse selection is relatively limited.
5Since the strength of interfirm ties vary, we can construct various measures for the quality of privateinformation a bank may possess about a firm dependent on who the bank lends to within that firm’sgroup, and the relative strength of those interfirm relationships.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 4
Karlan and Zinman’s novel experiment separately identified adverse selection and moral
hazard in microcredit but only found weak evidence for adverse selection effects. It is
unclear if the small effects they find are due to the small loans in the microcredit sector, or
because the experiment only used individuals who borrowed from the lender previously, or
if adverse selection is not a key problem in lending. Ausubel [1999] uses a randomized trial
which varied the contractual terms for a pre-approved credit card for 600,000 individuals.
Contrary to Karlan and Zinman, Ausubel finds large effects of adverse selection in the
credit card industry. To the best of our knowledge our paper is the first paper that
examines adverse selection in corporate lending. Filling the gap in the literature, we
examine the effects of adverse selection in a market with much larger loans and in a
setting where we expect the effect of adverse selection to be very different. Additionally,
we examine how lenders react to the change in a set of observable characteristics – which
is the key mechanism leading to credit market failure in the theoretical literature.
The classic literature on information problems in credit markets highlights the effect of
asymmetric information on symmetrically informed lenders. Recent papers examine the
theoretical and empirical effect of information asymmetries across lenders. Stroebel [2013]
theoretically models the interest rates a borrower receives when mortgage lenders are dif-
ferentially informed about the borrower’s collateral. He shows empirically that the return
is higher for more informed lenders. Moreover, due to the winner’s curse, less informed
lenders charge higher interest rates when competing against more informed lenders.
Consistent with Stroebel’s results we find the more informed lenders to be better off.
Yet our environment has richer intertemporal variation that allows us to address different
issues. First, our model endogenizes a firm’s total borrowings and we empirically test
whether the more informed lenders offer larger loan sizes than other lenders. Second, while
Stroebel relies on pre-existing differences in whether the lender also built the property,
we identify the effect of differential information exploiting an exogenous change in public
information reporting. Third, because firms in our setting have multiple loans, we can
estimate the effect of differential information across lenders for the same firm.
Public information performs three key roles in credit markets: (i) it reduces information
asymmetries between the borrower and the lender, (ii) reduces information asymmetries
between lenders and (iii) reduces a borrower’s incentive to default. However, public in-
formation does have drawbacks. First, it may reduce a lender’s incentive to procure
information if the information will be later revealed by a credit registry. Second, public
information may be imprecise or noisy causing excessive volatility in the observed public
information (Morris and Shin [2002]).
Public information is often considered a substitute for private information, or a leveler
of the playing field, but this is not always true. For example, consider a scenario where
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 5
two banks can lend to the same firm but only one bank knows the identity of all the
directors of this firm. Providing a directory of director creditworthiness to both firms,
would only benefit the firm who is privately informed about the firm’s composition of
directors. Therefore, it is possible for public information to be a complement to private
information.
Hertzberg et al. [2011] empirically demonstrate how merely the prospect of publicly an-
nouncing a bank’s private credit rating about a firm can lead to strategic effects by lenders,
and the subsequent reduction of credit. Lenders strategically reduced lending to borrow-
ers who they had previously labeled as a poor credit. They argue that lenders reduced
lending since the public information revelation would lead to other lenders reappraising
their lending terms and potentially reducing their credit lines to the borrower, causing the
borrower financial stress.
Our paper is similar to Hertzberg et al. in that we analyze the effects of altering public
information in an environment with multiple lenders. Since the change in public informa-
tion is common to all lenders, lenders must take into account how the change in public
information affects their willingness to lend and how the change affects their competitors’
willingness to lend.
Although both Hertzberg et al. and our paper examine changes in public information,
the form of the public information is quite different. Hertzberg et al. analyze a reform
which publicly released a lender’s appraisal of a firm’s creditworthiness whereas in the
reform that we analyze, the credit registry altered the information available about the
firm’s ongoing credit history. Hertzberg analyzes a reform where lenders respond to the
expectation of a bad credit rating becoming public knowledge limiting the firm’s capacity
to procure credit; by contrast, we study a reform where the SBP removed the ongoing
capacity to monitor a firm.
This paper contributes to the literature on relationship banking (Petersen and Rajan
[1994], Berger and Udell [1995], Degryse and Van Cayseele [2000]). The prior literature has
highlighted the potential for banking relationships to overcome the problem of information
problems. Our paper emphasizes that the strength of these prior relationships does not
completely eliminate the problem of imperfect information.
There is a nascent literature examining how variation in private information affects loan
officers’ decisions. Hertzberg et al. [2010], Cole et al. [2012], Paravisini and Schoar [2012]
all examine how loan officers evaluate whether they should offer loans, and at what terms,
under different information structures. This is similar to our paper’s environment except
the loan officers must consider not only what information they can observe, but what
other information may be available to other lenders.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 6
Our paper examines the importance of interfirm relationships in a firm’s ability to procure
credit. Khwaja and Mian [2005a, 2008], Khwaja et al. [2011] show two key and related
points: first, interfirm relationships allow insurance against idiosyncratic shocks; and sec-
ond, politically connected firms are able to garner political favors for the entire set of
related group firms. There is similar evidence in developed economies, Haselmann et al.
[2013] show how social connections between banks and firms can facilitate more favorable
lending terms for the firm. This paper emphasizes a mechanism for improved access to
credit that is distinct from the intragroup lending channels studied in previous work. In
particular, we suggest that the interfirm relationships may help firms procure loans since
a lender considers the creditworthiness of both the firm and the firm’s group. Therefore,
being part of a group with a perfect credit history can facilitate a firm’s access to credit.
To our knowledge, this is the first paper that shows that information about a firm’s group
has implications on a firm’s ability to borrow.
3. Data and Institutional Background
Bank lending is the primary source of formal funding in Pakistan. For instance in 2002,
Pakistan’s main stock exchange, the Karachi Stock Exchange, only had a market capital-
ization of 16% of GDP, which is much smaller than the more mature NYSE which had
a market capitalization of 92% of GDP (Khwaja and Mian [2005b]). The small size of
public equity and debt issuance is a common feature amongst emerging stock markets. In
part this is due to institutional failings, for example Khwaja and Mian [2005b] show that
brokers in Pakistan manipulated public stock prices through “pump and dump” schemes
to earn rates of return 50-90 percentage points higher than outside investors.
The data comes from the State Bank of Pakistan’s electronic Credit Information Bureau
(e-CIB), which legally requires all banks and lending institutions to submit data on all
borrowing firms with outstanding loan amounts greater than 500,000 Pakistani Rupees
(equivalent to about $8500 in 2004).6 Some of the information collected by the SBP was
passed back to the banks to facilitate lending. The information was provided through
“credit worthiness reports.” A sample report is shown in figure 20 in the Appendix.
The creditworthiness report provided information about the firm’s total borrowing, over-
due loans, ongoing litigation against the firm, and amounts written off in the last five
years. In addition, the central bank provided information on a firm’s group borrowing,
that is, the total borrowing by all firms which shared a mutual director, and whether the
group had any amount overdue. However, the central bank did not provide information
on which firms were in the firm’s group – it was solely the aggregate group’s credit history.
6This limit was removed in April 2006, and all loans (regardless of size) were required to be reported tothe e-CIB.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 7
Therefore, the individual banks would have different information about who – and who
was not – in the firm’s group.
Financial institutions use the credit reporting system as an initial appraisal and to monitor
the ongoing creditworthiness of a firm. In an interview a former loan officer remarked:
“The eCIB is used to verify credit history and monitor exposure, both during and after
approval [of the loan].” Additionally, the banks’ written notes on a firm’s credit worthiness
mention that credit reports were checked.
In April 2006, the central bank instituted new policies about the amount of information
they would provide about a firm through the credit reports. Specifically, the central
bank reduced the amount of information they would provide about a firm’s “group.” An
example of the new report is shown in Figure 21 in the Appendix. The main difference
between the reports is that the key terms detailing the group’s outstanding loans were
removed in 2006, hampering the lending institution’s ability to conduct due diligence on
the firm and group. A bank has limited capacity to recoup its funds on non-performing
loans – emphasizing the need to conduct sufficient due diligence prior to offering a loan.
In 2005, Pakistani banks recouped a mere 14.1% of the value of the loans which were
classified as Non-Performing Assets (Ministry of Finance [2006]).
Further there is important information in the creditworthiness of a firm’s group. The
probability of any corporate loan in our dataset being overdue in December 2004 was
2.6%. However, conditional on any firm in the firm’s group being overdue on a loan, that
firm was 6.2 percentage points more likely to be overdue on his loan. After controlling for
observables such as loan size, total borrowings, bank fixed effects, those firms who had an
overdue firm in their group were 6.8 percentage points more likely to be overdue on their
loans than firms with perfectly creditworthy partners. This highlights the importance of
understanding the creditworthiness the firm’s group in assessing the creditworthiness of
the firm.
This paper argues that this reduction in information had a major impact on bank lend-
ing decisions. It led banks to lend more to those firms about which they had private
information.
3.1. Why did the State Bank of Pakistan alter the information available to
lending institutions? Prior to April 2006, the State Bank of Pakistan defined a firm’s
group as those firms which shared a director. However, due to firms lobbying the SBP,
this definition of groups was altered, because the SBP believed that the definition of a
group was too broad and not an adequate measure of control.
The following quote comes from the minutes of an interview with Mr. Inayat Hussein,
head of the Banking and Regulation at the SBP — the group responsible for designing
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 8
and implementing the prudential regulation titled “Criteria of Grouping Companies for
the CIB Report,” and outlines the motivations behind the change:
“The 2004 criteria of grouping for the purpose of CIB reports resulted
in tying together some companies/firms/individuals, otherwise historically
financially sound, with defaulters. This happened due to common direc-
torship definition which also includes nominee directors who have little
influence on the management of the firm. Hence, the SBP management
decided to recognize and differentiate between those controlling shares in
a group from common directors having no influence over the management
of the group in question. Notwithstanding, SBP also appreciated financial
institution’s need to take informative decisions if complete information of
allied companies is not provided in the credit worthiness reports.”
These views were represented in the confidential internal minutes of a meeting of the SBP
in May 2004. The State Bank’s intention was to provide a new definition for groups – one
which would offer a better measure of control than the previous definition. However, the
SBP delegated the responsibility for constructing the group relations to the banks, stating
that “the onus for correct formation of the group as per definition given in the Prudential
Regulations will be on Banks” (State Bank of Pakistan, 2004 Prudential Regulation). So
the banks – the main beneficiaries of receiving information about a firm’s group – were
the ones expected to provide the information to the SBP.
The State Bank specifically warned against reporting too large a group, stating that:
“Banks are advised to be very careful while reporting the names of group entities in the
CIB data. In case any party disputes the group relationship, the reporting Bank should
be able to defend its position with documentary evidence” (State Bank of Pakistan, 2004
Prudential Regulation), further reducing the bank’s incentive to report a firm’s group to
the SBP.
Ultimately, the regulation led banks to report almost no group information to the central
bank. Since the SBP was no longer constructing group liabilities, and banks were not
reporting group entities, banks were left to construct their own groups. Therefore, they
were forced to make their own definition of groups and conduct their own due diligence.
A current credit officer remarked: “With the new e-CIB system, you have do your own
intel and also consider past information for firms” and in the new system “the full group
information was not declared – overall I preferred the older system.”
3.2. Other sources of firm information for a bank. Lenders access multiple different
sources of information about a borrower prior to offering a loan. They collect information
from the borrower using a standard form – the Basic Borrower Fact Sheet (BBFS), the
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 9
firm’s accounts, and some information held by the Securities and Exchange Commission
of Pakistan (SECP). Lenders also consult other people in the banking industry.7 The
information reported in the BBFS and SECP detail who are the firm’s directors and the
amount of shareholding each director holds.
Thus if a lender has a loan application from two firms, who share a mutual director, using
the information from the BBFS and the SECP, the lender can determine the existence of
an interfirm relationship and the strength of the relationship (as proxied via the share-
holdings). Further, we suspect if a bank lent to two members of the same group, the bank
would have better information about that firm’s group.
3.3. Building firm links. The State Bank of Pakistan collects information on all direc-
tors of a firm that borrows from a bank. This includes the director’s name, father’s name,
a common identifier, shareholding and home address. At the baseline of December 2004
we have details of 174,244 director relationships and 97,449 firms.8
3.4. What does the network of firms look like? Figure 1 shows the network of firm
connections across the entire set of borrowing firms in Pakistan in December 2004. A
connection between two firms is shown if both firms have at least one mutual director.
The most visually striking aspect of the network is the huge dense network of firm con-
nections in the center of the figure. The largest component9 is a total of 2395 firms in
December 2004.10
Figure 5 in section 8 shows what the network of firms looks like if we restrict attention
to interfirm relationships where a mutual director owns at least 25% of the firm’s equity.
Figures 6 and 7 show the distribution of the number of connections each firm has and the
distribution of component size.
7It was surprising to us that loan officers mentioned talking to other banks to learn more about a client’scredit worthiness. A further way to procure knowledge of a firm’s group is via the SBP, lenders can querywhich firms a director is part of, however, our interviews with loan officers suggest that the knowledge ofthis facility is not apparent.8The e-CIB is missing the common identifier for a total of 19,473 of these director relationships and theseare omitted from our sample. Since the common identifier was also used to compile which firms were ingroups, our definition of which firms were in groups should not be affected.9A component is defined as, the set of nodes such that every node within the component has either a director indirect connection to every other node within the component.10It should be noted, we are only building the component of firms using the set of firms currently bor-rowing. Therefore, the giant component should really be seen as a underestimate of the true size of thelargest component since there are possibly non-borrowing firms which would link other firms into the giantcomponent.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 10
Figure 1. The network of firm connections
Only firms with at least one connection are shown in the graph above. The circles andedges in the picture correspond to firms and connections between firms respectively,where a connection is defined if both firms have at least one mutual director. Those
firms with more connections are slightly darker and larger in the diagram.
3.5. The data. The loan level data comes from the SBP’s credit registry. The data
includes information on all directors for each firm and the amount of equity each director
holds. Furthermore, the data set in December 2004 included data on a total of 97,449
firms. 11,395 are corporate firms and 86,053 are sole proprietorship or small firm loans. In
April 2006, as part of an overhaul of the credit registry database, two separate registries
were created: (1) corporate firms and (2) all consumer. The data on sole proprietor loans
was moved into the consumer database and the data on corporate firms was kept within
the corporate database. Simultaneously, some of the original corporate firms were placed
in the consumer dataset – therefore, to maintain a consistent dataset of firms throughout
the period, we use the set of firms which were ex-post defined to be corporate firms and
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 11
maintained within the credit registry. Further, we exclude publicly listed firms due to our
expectation that there should not be effected by the reform and a lack of sufficient sample
size.
The SBP defined a firm’s group as the set of firms with a common director in the entire
borrowing data set of 97,449 firms. This paper follows the same definition of a group,
using the entire database of 97,449 firms in December 2004.
The data set stretches from December 2004 to December 2008, however there are data
validation concerns immediately following the regulation change. Therefore the months
from April to August 2006 are dropped from the data set. The new credit registry required
more information on borrowers to be uploaded by banks (for SBP’s role of supervising
of the banking system, not for the purposes of the credit registry11) and subsequently
required banks to upload the data in a new format which led to some initial teething
problems. In all our main within-firm regressions we include a bank interacted with
time fixed effect, therefore, if there was any time-varying bank-specific measurement error
(which was common across all the bank’s loans), then this fixed effect should alleviate this
concern.
In addition, one of the banks, Union Bank, was taken over in late 2006. Since data
reporting was poor during the takeover, that bank is omitted from all specifications.
The paper examines “funded” loan balances. A funded loan is a credit which is backed
by the bank, such that in the case of defaults, the bank must attempt to recover the loan
directly from that firm or person. The bank is the residual claimant on the loan. On the
other hand, non-funded loans are backed with a letter of credit or a personal guarantee
— if the borrower defaults, the bank can repossess funds from the guarantor.
The majority of lending are working capital balances, which are normally renewed every
12 months. They are similar to an overdraft facility, where firms are able to borrow more
or less at any stage subject to their total borrowing limit. Unfortunately, data on the
borrowing limit was not collected by the SBP prior to April 2006, therefore, the paper
restricts attention to the total amount borrowed in any month.
The data set details a loan to be overdue in any particular month if the loan amount was
overdue for more than 90 days.
There are a total of 94 banks which offer loans to corporate firms in 2004, but the sample
is restricted to banks offering at least 50 corporate loans in December 2004. This leaves a
final data set of 55 banks and 96.4% of the original data.
11It is important that the extra information being collected in 2006, is solely for purpose of SBP’s role ofsupervising the credit market since if the credit registry was displaying more information than it was in2004, this could potentially conflate some of the observed effects of the regulation change.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 12
The sample is restricted to both harmonize the data set and for tractability. Many of the
specifications use three high dimensional fixed effects: ‘firm interacted with date’, ‘firm
interacted with bank’ and ‘bank interacted with date’. Therefore, by removing the very
smallest banks, the computation is greatly sped up.
All variables which include nominal amounts have been discounted to December 2004
prices using the official Pakistani CPI index published by the State Bank of Pakistan.
Tables 1 and 2 present summary statistics on the entire dataset as well as showing more
detailed statistics on group firms. Figures 8 and 9 show the distribution of total firm
borrowings and the distribution of loan sizes for group firms.
4. Model
The model demonstrates how lenders who are differentially informed about a firm’s cred-
itworthiness decide whom to lend to, and how much to lend, after assessing the creditwor-
thiness of borrowing firms and what information other lenders may have.
The model aims to show how lending patterns change between informed and uninformed
lenders as we alter the composition of the borrower pool, and as we alter the informational
differential between lenders.
4.1. Setup. There are three players in the model; two lenders and a single firm. There
are two different types of firms, a high type, H, and a low type, L, which vary in the
probability of repayment. The probability of the low type is γ and the probability of the
high type is (1− γ).
Both lenders have the same cost of capital (ρ), however one of the lenders (I) – the informed
lender – is more informed of the firm’s type than the other (N) – the uninformed lender.
The firm has no outside source of funds, no collateral and limited liability. The firm is
able to undertake a project such that if a firm of type i invests k in the project, the firm’s
output Yi(k) is:
Yi(k) =
Ak with probability Xi1+k
0 else
Where XL < XH < 1, therefore the high type is more likely to have a successful project
conditional on the amount borrowed, k. The project’s expected output, E[Yi(k)] is in-
creasing and concave in capital, so the expected return E[Yi(k)/k] is decreasing in capital.
For simplicity, we shall assume that the interest rate (R) is exogenously fixed at a rate
greater than the cost of capital (ρ) and lower than the return of the project (A) if successful:
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 13
ρ < R < A
This is not an innocuous assumption since it stops competition over interest rates. Yet it
seems empirically plausible as Khwaja and Mian [2008] demonstrate for Pakistani corpo-
rate firms and Petersen and Rajan [1994] for small business firms in the US, interest rates
are not responsive to changes in lending costs or information respectively. Both papers
find large effects on the amount of credit each bank is willing to offer.
To ensure there is the prospect of there being “lemons” in the model we assume:
(1)RXL
ρ< 1 <
RXH
ρ
Inequality (1) states the maximum expected return from the low (high) type borrower is
lower (greater) than the cost of capital for the lender. Therefore inequality (1), states that
if the lenders knew the firm was a low-type, it would not be profitable to offer that firm
any loans.
Further, for ease of exposition in the proofs12 we assume that RXLρ > 1
2 and we assume
that RXHρ < 4.
We assume a lender can only make non-negative loan offers, kj ≥ 0.
The firm has limited liability, no outside wealth and no collateral. Further, the firm cannot
strategically default on a loan – this could be due to legal requirements or the lender can
repossess the firm’s output. We define a firm default at lender i, Di, if a firm received a
loan and the project was unsuccessful.
For simplicity we have assumed that both the informed lender and the uninformed lender
know that there exists one lender of each type. In the context of the empirical setting this
is not so clear. Banks do not know what information other banks do and do not possess.
4.2. Timing.
(1) Nature chooses the firm’s type Xi.
(2) The informed lender observes Xi.
(3) The informed and uninformed lenders make simultaneous bids (kj) over how much
to lend at an interest rate R.
(4) The firm accepts none, one, or both of the loan offers.
(5) The project is successful, or not, and payoffs are assigned.
12 These bounds on the set of parameter values are sufficient conditions which greater simplify the proofsfor certain boundary cases (when the uninformed lender does not enter).
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 14
Figure 2 outlines the extensive form of the game.
Figure 2. Extensive form of the game.
4.3. The game. We will consider the Perfect Bayesian Equilibrium of the game.
To solve for the equilibrium of the game, we will use backward induction. Firstly we solve
for the contract offers the firm will accept. Second, conditional on the firm’s strategy, we
solve for the optimal loan offer by each of the lenders.
4.3.1. The firm’s problem. The firm wants to maximise its expected utility, which takes
the form:
Ui(k,R) = Pr(success)× (Ak −Rk)
=Xi
1 + k(A−R)k
Where k is the firm’s total borrowings. This specific utility function has some key advan-
tages, which will greatly simplify the model. The firm’s utility Ui(k,R) exhibits strictly
increasing returns in capital for all interest rates below the productivity parameter, A.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 15
Therefore, the firm’s weakly dominant strategy will be to accept all loan offers.
4.3.2. The informed (I) lender’s problem. Having solved the firm’s problem, we know the
firm will accept all loan offers. The informed lender’s problem becomes:
πiI(kiI , kN , Xi) = max
kix
Pr(success)× (kiI)− ρkiI
=
[Xi
1 + kiI + kNR− ρ
]kiI
πiI(kiI , kN , X) ≥ 0
kiI ≥ 0
Where kiI is the informed lender’s loan offer to the firm of type i and kN is the uninformed
lender’s loan offer. The informed lender’s problem is to maximise the expected return
from lending to a borrower (conditional on the uninformed lender’s loan amount) minus
the cost of lending.
Recalling the assumption that the cost of capital is greater than the maximum repayment
from the low type (ρ > XLR). In this case, the informed lender will not lend to the
low-type firm, since the expected profit from any non-zero capital offer is negative:
πLI (kLI , kN , X) < 0 ∀kLI > 0
Therefore, the informed lender will make the offer kLI = 0 in equilibrium.
4.3.3. The uninformed (N) lender’s problem.
E (πN (kI , kN , X)) = maxkN
EX {Pr(success)× (RkN )− ρkN}
= EX
{[Xi
1 + kiI(Xi) + kNR− ρ
]kN
}E (π(kI , kN , X)) ≥ 0
kN ≥ 0
The uninformed lender’s problem is similar to the informed lender’s problem. However
the uninformed lender does not observe the firm’s type, so the lender must maximise over
the expectation of the firm’s type. It should be noted that the uninformed lender must
also consider that the informed lender’s loan offer will be a function of the firm’s type. In
particular, the uninformed lender will face more competition on the high-type firms than
the low-type.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 16
4.4. Equilibrium.
Proposition 1. A Perfect Bayesian equilibrium of the game is:
(informed lender) kLI = 0, kHI = kH∗I > 0
(uninformed lender) kN = k∗N ≥ 0
(firm) sif = (Acc, Acc)
Proof. In the appendix. �
Therefore, the informed lender will always offer the null offer kLI = 0 to the low-type firm.
Further the informed lender, will always make a positive loan offer (kH∗I ) to the high-type
firm.
The uninformed lender will make an offer an offer k∗N , which may be the null offer. The
uninformed lender makes the null offer k∗N = 0 when the problem of adverse selection
is sufficiently severe that the uninformed lender is unable to make positive profits when
entering the market.
The firm’s weakly dominant strategy is to accept all loan offers13.
Proposition 2. The informed lender will lend more in expectation than the uninformed
lender:
∆k ≡ γkLI + (1− γ)kHI − kN > 0
Proof. In the appendix. �
Proposition 2 shows that the informed lender will make larger loans on average than the
uniformed lender. If the firm is a high-type, the uninformed lender competes with the
informed lender on offering a loan, therefore, they split the profits from servicing the
high-type firm. However, on the low-type firms, the uninformed lender is the sole provider
of loans and makes a loss. Overall, the losses from the low-type firm leads the uninformed
lender to make smaller loans on average.
Proposition 3. If the uninformed continues to lend, the uninformed lender will have
greater rates of default than the informed lender:
∆D ≡ E[DN −DI |kN > 0
]> 0
13There is another PBE where both the informed and uninformed lender’s make the null offers kLI = kH
I =kN = 0, and the firm rejects all loan offers. Given these strategies, no lender or firm could be better off.However, we ignore this equilibrium in our analysis, since it involves the firm playing a weakly dominatedstrategy.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 17
Proof. In the appendix. �
The information asymmetry leads the uninformed lender to offer loans to the low-type
borrowers leading to greater default rates. Combining proposition’s 2 and 3 the uninformed
lender lends less than the informed lender and makes worse quality loans.
The lack of public information leads to two sources of welfare loss: (1) the good borrowers
can receive too much capital and (2) the low-type borrowers receive too much capital.
Proposition 4. As we reduce XL, the expected difference between how much the informed
and uniformed lender offer is increasing.
∆ ≡ (1− γ)kI − kN is decreasing in XL
Proof. In the appendix. �
As we decrease the quality of the “lemons” in the model, the uninformed lender makes
greater losses by servicing the entire market. This leads to the uninformed lender reducing
her overall lending.
Since the informed lender never makes an offer to the low-type borrower, the informed
lender is affected solely through the reduction in the uninformed lender’s willingness to
offer loans. If the uninformed lender reduces the size of her loan offer, the informed lender’s
optimal reaction is actually to increase her offer.
4.5. Allowing for differing quality of information. We can extend the model to allow
for varying differences in information between the informed lender and the uninformed
lender.
Formally, instead of the informed lender knowing the borrower’s true type, let us as-
sume the informed lender receives a signal X̃ = {x̃L, x̃H} about the borrower’s type. In
particular:
Pr(Xi = XL|x̃L) = q >1
2Therefore, in this version of the model, the informed lender still has some uncertainty over
the firm’s type.
Then, let us assume that the informed lender does not want to lend if she observes the
low signal x̃L, formally:
R [XLq +XH(1− q)] < ρ < R [XL(1− q) +XHq]
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 18
This assumption is the analogue of equation (1) in the basic model. This equation states
that if the informed lender receives a signal x̃L the maximum expected return is strictly
less than the cost of capital.
Altering the timing of the model slightly:
(1) Nature chooses the firm’s type Xi.
(2) The informed lender observes X̃i.
(3) The informed and uninformed lenders make simultaneous bids (kj) over how much
to lend at an interest rate R.
(4) The firm accepts none, one, or both of the loan offers.
(5) The project is successful, or not, and payoffs are assigned.
4.5.1. Equilibrium.
Corollary 1. A Perfect Bayesian equilibrium of the game is:
(Informed lender) k̃LI = 0, k̃HI = k̃H∗I > 0
(Uninformed lender) kN = k∗N ≥ 0
(Firm) sif = (Acc, Acc)
Proof. The proof of this is very similar to the proof of Proposition 1. The details are in
the appendix. �
Proposition 5. As the informed lender’s signal of the borrower improves, the difference
in expected lending increases. Formally:
∆̃ ≡ [(1− q)γ + q(1− γ)] k̃I − kN is increasing in q
Proof. The proof of this is very similar to the proof of Proposition 4. The details are in
the appendix. �
The intuition for Proposition 5 follows from: As the informed lender’s information becomes
better, the informed lender is better able to target the good firms, therefore, the informed
lender is more willing to offer bigger loans. This leads to the uninformed lender making
smaller profits on the good borrowers (more competition), and subsequently reducing the
size of her loan offers.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 19
4.6. Mapping the model to the data. Empirical predictions of proposition 2: The
informed lenders will offer larger loans on average to those firms for whom they have
better information following the reform.
Empirical predictions of proposition 3: The informed lenders will have default rates similar
to the uninformed lender following the reform on those firms they both continue to serve,
but the uninformed lender will have greater overdue rates overall.
Empirical predictions of proposition 4: The informed lenders will lend more to those firms
where the quantity of lemons is greatest, since determining which firms are good credit
risks, and which are not, is more important as we increase the proportion of bad types.
Generally, those firms which have had an amount overdue in the past would be the riskiest
firms. We would expect the effect of the reform to be the largest on those firms.
Empirical predictions of proposition 5: As we increase the information advantage for the
informed lender, the expected difference in loan sizes will increase. We will proxy for
effects of the informational advantage by examining the effects of the reform on those
lenders who lent to a greater number of a firm’s group, and by carefully examining the
strength of the interfirm relationships between firms.
5. The Effect of the Reform on a Firm’s Source of Credit
5.1. Econometric specification. The paper’s main question is: What is the effect of
asymmetric information between lenders on a firm’s source of credit? We answer this
question using a reform that exogenously reduced banks information about firms in a way
that varied across bank-firm lending pairs. To test the model’s predictions, we examine
whether those banks with private information about a firm’s group were more likely than
other banks to renew a firm’s loans following the reform.
In the paper’s main specifications, a “group” is defined using a relatively narrow definition:
An overlapping director must own a substantial amount of equity in two firms before those
two firms are grouped together. We will demonstrate that altering the definition of a group
in economically meaningful ways will lead to different results.
Definition 1. Firm f ’s group is all other firms with whom firm f has a common director
and at least 25% shareholding in both firms as of December 2004.14
The main source of identification in the paper will be to compare borrowing amounts
for the same firm from two different lenders, before and after the regulation, where each
lender lends to different members in the group at baseline (so we are restricting the sample
14This definition of groups allows firms to be in multiple groups and that groups are not mutually exclusive.Therefore, a group is defined with respect to a firm. For a quick pictorial representation how firms can bein multiple groups see figure 4.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 20
Figure 3. Identification Strategy: comparing loan details for the samefirm borrowing from two different banks.
to lenders who were already lending before the reform, and therefore, by definition have
some information). For example, assume there are two firms “F1” and “F2” who are in
the same group. Both “F1” and “F2” borrow from two banks each in December 2004.
“F1” borrows from banks “A” and “B” and “F2” borrows from banks “B” and “C,” as
shown in Figure 3.
Notice that bank “B” is lending to both firms in the group in December 2004, whereas
bank “A” and bank “C” only lend to one firm in the group. After the regulation change,
only bank “B” is able to compile the group’s lending.
Definition 2. A loan from bank b to firm f is labeled a “informed loan” if at least
one other member of firm f ’s group borrows from the same bank, b, in December 2004
(baseline). Similarly, a loan from bank b to firm f is labeled an “uninformed loan” if no
other member of firm f ’s group borrows from the same bank, b, in December 2004.
In the stylized example in figure 3, firm F1 has one informed loan (the loan with bank
“B”) and one uninformed loan (the loan with bank “A”). The paper’s identification will
be examining how lending changes between the informed and uninformed loans, before
and after the regulation change.
We estimate equations of the form:
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 21
Ybft = abf + abt + aft + β1 × postt × Informedbf + εbft
The unit of observation is at the bank-firm-date level, so Ybft is the variable of interest
at bank b, firm f in month t. For example, it could be the size of the loan outstanding
by firm f at bank b, in month t. Informedbf is a dummy variable equal to one if the loan
between bank b and firm f is an informed loan. Post is a dummy variable equal to one
for a loan after April 2006.
The standard errors εbft are clustered at the level of the component – at the level such
that every firm within the component has at least one indirect link to every other firm in
the component.15
All the main regressions contain a firm×date fixed effect, αft. This fixed effect implies
that we are estimating the difference in firm lending using differences for the same firm
and in the same month. Therefore, we are estimating β1 from only those firms that have
both an informed and an uninformed loan. We have a total of 449 firms and 1,784 loans in
December 2004 who had both an informed and an uninformed lending relationship. Table
3 presents more details about the firms which identify β1.
Figure 11 demonstrates that the firms who identify β1 in general have larger loans. This
is in part mechanical, since we are only identifying the effect of the policy from those firms
who have at least two loans.
Including the fixed effect abt ensures that we are allowing for any aggregate change in
bank lending for each month, and the fixed effect abf ensures we are controlling for any
firm-bank specificity.
This section restricts attention to those banks that were lending in December 2004, there-
fore, the paper does not include any new relationship in the analysis because any new
borrowing relationship may be endogenous. Furthermore, in the estimation procedure,
firms who discontinue relationships with all of their original lenders are dropped16.
5.1.1. Outcomes of interest. There are three main outcomes of interest in the paper:
• Log loan sizebft
• Renewed loanbft
15A component is defined as the set of firms for whom their exists a direct or indirect link to every otherfirm within the component. A direct link between two firms is defined if there is at least one mutualdirector who owns at least 25% of each firm and an indirect link exists between two firms if there is atleast one path of direct links between the two firms. Conceptually, if firm f is in component c then firmf ’s group must be a subset of a component c. The definition of a link, is similar to the one used in allbenchmark specifications.16The paper shows robustness results showing similar effects with the set of firm-bank connections whichalways have an active borrowing relationship.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 22
• Overduebft
“Log loan sizebft” is defined as the log of real funded loan size outstanding in date t at
bank b by firm f . If there is no loan observed this is coded in the data as equal to the
minimum of what is observable in the data set (the log of 500,000 Rs.).
“Renewed loanbft” is defined as whether firm f at bank b at date t has an outstanding
funded loan amount above 500,000 Rs. (in 2004 Pakistani Rupees).
“Overduebft” is defined as whether firm f at bank b at date t is overdue at date t. There
are certain endogeneity issues when looking at overdue rates because a firm can only
go overdue if a firm has a loan which will be shown to be a function of the amount of
information a bank has about a borrower. This is discussed in greater detail in the results.
5.2. Results. We explore the implications of differentially informed lenders on a firm’s
source of credit in three ways: (i) to explore the overall effects of the reform, we estimate
the difference in loan sizes between informed and uninformed lenders (ii) to explore the
role of private information, we estimate how the difference in loan sizes changes as we use
various measures of informed lenders and (iii) to explore whether there were heterogeneous
effects for different firms, we estimate the effect of the reform on different firm sizes and
differing credit risks.
5.2.1. Following the reform, loan sizes were relatively larger within informed banking rela-
tionships. Our main results indicate that when public information available to banks was
reduced, the change caused those banks to lend more to firms for whom they had greater
private information. Banks with a better knowledge of a firm’s group were more likely to
renew the credit facility, and grant greater credit for renewed loans.
This is most strikingly represented in Figure 12. Figure 12 plots the coefficients from a
regression of loan renewal on an informed dummy, interacted with date dummies, and
all second-order fixed effects, firm×date, date×bank, and firm×bank. The figure clearly
shows that the difference in renewal rates between informed and uninformed loans is
relatively constant before the reform, suggesting that the common trends identifying as-
sumption holds. Following the reform, there is a sharp and persistent increase in the
renewals of informed loans relative to uninformed loans. This indicates that the reform
causes banks to increase their lending to firms for which they have more information.
Table 4 shows the estimates for the policy’s effect under different specifications. The
first column contains date, firm and bank fixed effects, and therefore is being estimated
from between- and within-firm differences. Columns 2-6 all include a firm interacted with
date fixed effects. This ensures that we only use the set of borrowers who have both, an
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 23
informed and uninformed banking relationship, to identify whether an informed banking
relationship was more likely to be renewed.
The estimates are all relatively similar and the various specifications suggest that an
informed banking relationship was between 5-8% more likely to be renewed than an un-
informed banking relationship for the same firm. The preferred specification is column 6,
which includes all three second order-fixed effects. In doing so, we are controlling for any
aggregate changes in a bank’s willingness to offer credit over time and firm-bank match
specificity, and identifying the effect from firms which had both, an informed and unin-
formed lending relationship. Figure 12 is the graphical counterpart of the regression in
column 6, except that we interact date dummies with the informed dummy variable.
While the previous results examine the effects of the policy on the extensive margin, here
we can examine if the total size of the funded loan was larger in an informed banking
relationship after the regulation change.
Figure 13 plots the coefficients from the log loan size on an informed dummy variable
interacted with date dummies and all second-order fixed effects.17 Though the estimates
are less precise than the extensive margin, we clearly see a similar trend as the extensive
margin. Table 5 shows the same specifications as Table 4 but the dependent variable is
the log of real funded loan size. The various estimates suggest that an informed banking
relationship was between 8-12% larger following the policy change.
As a robustness check, we restrict the sample to those lending relationships which last until
June 2008 or December 2008, and compare loan balances across informed and uninformed
banking relationships for the same firm. Table 6 shows the estimates from this regres-
sion. Although power is an issue, the size of the effects look similar across the different
specifications to the results in table 5.
The SBP only required banks to report details of a loan if the firm’s total loan outstanding
was greater than 500,000 Rupees ($8,500) could lead to the following bias in our results.
Loans which were initially just above the cutoff could be partially repaid, subsequently
falling below the 500,000 Rupees threshold and as such incorrectly categorized as a non-
renewed loan. This is a greater concern since the uninformed loans are in general smaller
than the informed loans. Table 7 excludes firms which had a loan close to the cutoff in
December 2004. The observed effect of the policy seems consistent when we exclude firms
with a loans below $12,750, $17,000 or $21,250 in December 2004. This suggests that the
censoring of the data at 500,000 Rupees does not affect the results.
17If a loan is not observed, we code the loan size to be the log of 500,000 Rs., which is the minimumthreshold at which we are able to observe a loan.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 24
5.2.2. There were substantial differences in the effect of the policy on the credit market
depending on the strength of the interfirm relationship. The results in section 5.2.1 demon-
strate that the reform led banks with private information about a group to lend more than
other banks. Here we consider whether the effects vary by the amount of private informa-
tion an informed lender has. In particular we consider variation in: the strength of the
observed interfirm relationship and the number of firms who borrow from the same lender.
We observe substantial heterogeneity in the measured effect of the reform depending on
the level of control implied within the interfirm relationship. If we separate the interfirm
relationships according to the amount of equity a director owns, it is clear there was little
or no effect on those firms which had only overlapping directors. Further, the size of
the effect was increasing in the interfirm relationship according to equity levels, which is
a proxy for control. Therefore, this suggests that merely overlapping directors had no
informational content on a firm’s creditworthiness.
Table 8 and figure 14 shows the effect of the reform for different levels of equity held by a
mutual director. If a bank observed two firms who shared a director who owned at least
40 percent of both firms, the bank was 7.6% more likely to renew a loan from either firm,
compared to a bank who only observed one firm. However if a director had no ownership
stake in either firm, the bank was only 2.5% (not statistically significant) more likely to
renew their loans.
These results are consistent with proposition 5: As the information differential between
informed and uninformed lenders increases, the uninformed lender will lend less.
An additional measure of the strength of the firm-bank relationship is observing the num-
ber of other group firms who borrow from the same bank. If more firms in the same group
borrow from a bank, that lender is expected to have more private information about inter-
firm links, and therefore be more willing to lend. Tables 9 and figure 15 demonstrate that
as we increase the number of interfirm relationships borrowing from the same bank, the
effect of the policy was much larger. Those bank-firm pairs where the bank had greater
information over the group are observed to increase their lending more.
5.2.3. The effect of the change in information was predominantly felt by small to medium
sized firms. From a policy perspective it is important to examine what type of firms were
most affected by the reform. We examine if the effect varies by firm size.
Table 10 and figure 16 show the effects of the policy on the likelihood of a loan being
renewed by different deciles, where the deciles are created according to the total amount
the firm borrows at baseline. We see almost no effect (and certainly no statistically
significant effect) on the largest decile of borrowers.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 25
These results suggest that public information was most important for small to medium
sized firms. These results can be interpreted in two ways: (i) the relative cost of procuring
information is largest for the smallest firms or (ii) information about the largest firm’s
groups is already well known by the banks. We consider each possibility in turn.
As discussed in section 3.2, banks procure more information about a firm in addition to
what is provided by the SBP’s credit registry service. This information — such as calling
other bank managers, or accessing the SECP database — may be costly to acquire. It is
plausible that loan officers will conduct greater scrutiny over larger loans (assuming the
cost of default is linear in loan size). Consequently the effect of the regulation change
would be largest on those firms that borrowed the least (as this is where the proportional
cost of acquiring more information to dollar lent is the largest).
Mr. Mansoor Siddiqi, the ex-Director Banking Policy and Regulation Department at the
SBP remarked in an interview: “The number of corporate [firms] is not large in Pakistan
therefore people tend to know about reputations,” further strengthening the assertion
that we would expect little to no effect on those firms that had large groups. Table 11
and figure 11 show the effects of the policy on the likelihood of a loan being renewed by
different deciles, where the deciles are created according to the total amount the firm’s
group borrows at baseline. Similar to the results in table 10, we see no effect on those
firms who are part of the largest groups.
5.2.4. The effect of the policy was largest on those firms who were overdue on a loan in
December 2004. Another way to analyze how the reform may have had heterogeneous
effects across firms is to consider those firms who were observably worse credit risks prior
to the reform. Firms that had loans overdue in December 2004 are expected to be at
the greatest risk of being overdue in the future. Further, the problem of incomplete
information on the firm’s group may be more pronounced, since one firm with an amount
overdue might indicate financial distress within the wider group. So, a lending institution
may be more willing to lend to a firm if it is able to inspect the wider set of group firms
too. Also, when one firm defaults, the institution could react by reducing lending to the
group at large.
Therefore, a firm that has an amount overdue may be more likely to borrow more from
an informed banking relationship after the regulation change. Confirming this economic
intuition, these results are shown in table 12.
These results are consistent with proposition 4 in the model: In environments where there
is largest number of lemons, the uninformed lender is relatively more likely to stop lending.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 26
5.2.5. Overdue rates are similar across informed and uninformed banking relationships.
The previous results highlighted how the reform affected who was able to borrow. In this
section we examine if the reform affected overdue rates.
The reduction in public information may alter a firm’s incentive to default on a loan for
two reasons. First, a firm may be more willing to be overdue because the impact on the
rest of the firm’s group will be limited. Second, the lender would be less willing to offer a
loan. In section 5.2.1 we demonstrated that the uninformed lender was 5.4% less likely to
renew a firm’s loan than the informed lender.
Only if all loans were renewed would we be able to identify the effect of the change
information on a firm’s incentive to default.
However, if strategic default was a key problem in these corporate markets you would
expect the relative likelihood of a loan being overdue to be greater at an uninformed lender
than an informed lender for the same firm. The firm is less likely to have his loan renewed
by the uninformed lender, reducing the dynamic incentive to repay the loan. Also, the
repercussions on a firm’s group would be smaller at an uninformed lender. Therefore, the
absence of any differential on the likelihood of a loan being overdue between an informed
and uninformed lender suggests there was no greater strategic default by the firm upon
the uninformed lender.
The regressions in column 2 of Table 13, show little difference in overdue rates between
informed and uninformed banking relationships. Column 2, which estimates the difference
in the likelihood of a loan being overdue at an informed and uninformed lender for the
same firm, shows a relatively precise zero estimate. Column 2 suggests there was no
strategic default by the firm following the regulation reform – even though they were less
likely to have their loan renewed.
Column 1 is estimated from both, between-firm and within-firm differences, since it does
not include a firm×date fixed effect. Column 1 suggests uninformed lenders select a worse
set of loans to renew than informed lenders which is consistent with the proposition 3.
In particular, defaults rates for uninformed lenders were 3% higher than informed lenders
following the reform.
6. The Effect of the Reform on a Firm’s Access to Total Credit
In the previous section we established that a firm was more likely to receive a loan from
an informed lender following the reform. In this section we wish to examine whether those
firms with such informed lenders were more likely to have larger credit lines following the
reform. Consequently, we ask: Did firms merely substitute their lending partners and
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 27
receive the same total loan amounts? Or, did the reform lead to real effects in how much
a firm was able to borrow?
6.1. Econometric specification.
Definition 3. Firm f has an informed lending relationship if firm f borrows from a bank
who also lends to at least one other member of firm f ’s group in December 2004.
To examine whether the reform affected a firm’s access to credit we create a dummy
variable “informed lender” which takes a value of “ 1” if the firm had an informed lending
relationship in December 2004.
We identify the effect of the reform by comparing total loan amounts for those firms with
and without an informed lending relationship before and after the regulation. Formally,
we estimate an equation of the form:
Yft = af + at + β1 × postt × Informed Lenderf + γXft+εft
Where Yft is the log of real total funded borrowings by firm f in month t. Informed lenderf
is a dummy variable equal to one, if the firm has at least one informed lending relationship
in December 2004. Post is a dummy variable equal to one for a loan after April 2006.
af and at are firm and date fixed effects respectively. In section (5) our identification
relied on comparing the loan outcomes for the same firm, whereas in this section we rely
on comparing the total credit borrowed between firms. Consequently, to allow for firm
differences across regions and sectors, we include a set of firm by time controls Xft, such
as the firm’s province interacted with time fixed effects.
As before, the standard errors εft are clustered at the level of the component – at the level
such that every firm within the component has at least one indirect link to every other
firm in the component. As in the previous section, we do not observe loans smaller than
500,000 Rupees. Therefore, if we do not observe any loan for a firm, we code that firm’s
total borrowings to be 500,000 Rupees.
6.2. Results.
6.2.1. Firms with an informed lending relationship borrowed more after the reform. Table
14 estimates the effect of the reform on firms who had an informed lending relationship
in December 2004. The difference-in-difference estimates suggest that those firms with
an informed lender were able to borrow 11%-14% more following the reform, compared
to firms with no informed lender. The results are robust to various controls such as a
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 28
province interacted with date fixed effect and business sector interacted with date fixed
effect.18
Combining the results from section 5.2.1 with the results in table 14 suggests that the
reform had two key effects: (i) lenders with greater information lent more to the same
firm following the reform and (ii) those firms with informed lenders were able to access
more credit following the reform.
This suggests that the reform not only led to a reallocation of lending across lenders but
also led to real effects in the credit market by altering a firm’s capacity to procure credit.
Interpreting these results, the reduction in public information led to a reallocation of credit
to borrowers for whom the lender had greater information. Ultimately, this disadvantaged
firms who borrowed from banks which had limited information about their wider set of
interfirm relationships.
6.2.2. The borrowing capacity of the smallest firms were most affected by the reform. The
effect of the reform was greatest on the smallest firms’ total borrowings. We have estab-
lished that the reform had real effects on how much a firm was able to borrow. To further
evaluate the welfare and policy implications, we ask: What type of firms were the most
affected by the reform?
To answer this question we estimate the effect of the reform on different sized firms.
Table 15 estimates the effect of the reform on different deciles of firm size and whether
they had an informed lending relationship in December 2004. The difference-in-difference
estimates suggest that those firms in the smallest decile were the most affected by the
reform.
Combining the results from section 5.2.3 suggests that public information was crucial
for the smallest firms. Not only were the more informed lenders more likely to lend
to the smallest firms, but the smallest firms with an informed loan were more likely to
procure larger total borrowings. The change in total borrowing for the largest decile was
not significantly different from zero. These results further support the evidence that the
regulatory reform had significantly different effects for different sized firms. With the
largest effects for the smallest firms.
7. Conclusion
Asymmetric information is a focal issue when studying credit markets. These information
asymmetries may exist between a borrower and a lender, and between different lenders.
18It should be noted the although the results are robust to a group size specific time trend, the results arenot robust to a group size fixed effect interacted with a date fixed effect.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 29
To investigate the implications of adverse selection among differentially informed lenders,
we use a reform by the State Bank of Pakistan.
In April 2006, the State Bank of Pakistan instituted a credit registry reform which exoge-
nously reduced public information about a firm’s creditworthiness, and did so in a way
that varied across firm-bank pairs.
First, we present a model of differentially informed lenders with adverse selection, which
generates clear predictions on how the reform is expected to affect a bank’s willingness to
lend. The model predicts that those banks with greater private information about a firm
would be more likely to continue to lend to that firm. Further, it may lead those lenders
with the least information about a firm to stop lending.
Then, to test the model’s predictions we utilize the natural experiment that the SBP’s
reform generated. We show empirically that the absence of public disclosure of group
affiliation led to a reallocation of corporate borrowing. Those lenders who had greater
private information were 5.4% more likely to renew a borrower’s loan. Remarkably, this is
true even for banks that had a pre-existing relationship with the firm, suggesting that the
strength of prior relationships does not eliminate the problem of imperfect information.
Not only do we see a reallocation of corporate borrowing, we also demonstrate that those
firms who borrowed from an informed lender were able to borrow 12.3% more than firms
who borrowed from less informed lenders. Consequently, we show that the source of a
firm’s credit, as well as the total credit it could procure, were affected by the State Bank
of Pakistan’s reform.
The model suggests that the reduction in public information leads to a negative welfare
effect. Firstly, the asymmetry of information reduces the contestability of the market, and
secondly, the reduction in information amplifies adverse selection, subsequently increasing
the misallocation of credit. Finally, when the severity of adverse selection becomes worst,
the uninformed lenders may stop lending.
Ultimately, complex interfirm relationships can alter a borrower’s incentives and ability to
repay a loan – necessitating that a lender analyze the firm’s creditworthiness and consider
the firm’s interfirm relationships. Public information about these relationships is key to
reducing information asymmetries, both, between a borrower and a lender, and between
lenders themselves.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 30
References
G. A. Akerlof. The market for ”lemons”: Quality uncertainty and the market mechanism.
The Quarterly Journal of Economics, pages 488–500, 1970. URL http://www.jstor.
org/stable/10.2307/1879431.
H. V. Almeida and D. Wolfenzon. A theory of pyramidal ownership and family business
groups. The Journal of Finance, 61(6):2637–2680, 2006. URL http://onlinelibrary.
wiley.com/doi/10.1111/j.1540-6261.2006.01001.x/full.
L. M. Ausubel. Adverse selection in the credit card market. mimeo, 1999.
A. N. Berger and G. F. Udell. Relationship lending and lines of credit in small firm finance.
Journal of business, pages 351–381, 1995.
M. Bertrand, P. Mehta, and S. Mullainathan. Ferreting out tunneling: An application
to indian business groups. The Quarterly Journal of Economics, 117(1):121–148, 2002.
URL http://qje.oxfordjournals.org/content/117/1/121.short.
S. Cole, M. Kanz, and L. Klapper. Incentivizing calculated risk-taking: Evidence from
an experiment with commercial bank loan officers. Harvard Business School Finance
Working Paper, (13-001), 2012. URL http://papers.ssrn.com/sol3/papers.cfm?
abstract_id=2101648.
D. De Meza and D. C. Webb. Too much investment: A problem of asymmetric in-
formation. The Quarterly Journal of Economics, 102(2):281–292, 1987. URL http:
//qje.oxfordjournals.org/content/102/2/281.short.
H. Degryse and P. Van Cayseele. Relationship lending within a bank-based system: Ev-
idence from european small business data. Journal of Financial Intermediation, 9(1):
90–109, 2000.
M. A. Desai and D. Dharmapala. Corporate tax avoidance and firm value. The Review of
Economics and Statistics, 91:537–546, 2009.
R. Gopalan, V. Nanda, and A. Seru. Affiliated firms and financial support: Evidence from
indian business groups. Journal of Financial Economics, 86(3):759–795, 2007.
R. Haselmann, D. Schoenherr, and V. Vig. Lending in social networks. 2013.
A. Hertzberg, J. Liberti, and D. Paravisini. Information and incentives inside the
firm: Evidence from loan officer rotation. The Journal of Finance, 65(3):795–828,
2010. URL http://onlinelibrary.wiley.com/doi/10.1111/j.1540-6261.2010.
01553.x/full.
A. Hertzberg, J. Liberti, and D. Paravisini. Public information and coordination: Ev-
idence from a credit registry expansion. The Journal of Finance, 66(2):379–412,
2011. URL http://onlinelibrary.wiley.com/doi/10.1111/j.1540-6261.2010.
01637.x/full.
D. M. Jaffee and T. Russell. Imperfect information, uncertainty, and credit ra-
tioning. The Quarterly Journal of Economics, 90(4):651–666, 1976. URL http:
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 31
//qje.oxfordjournals.org/content/90/4/651.short.
D. Karlan and J. Zinman. Observing unobservables: Identifying information asymmetries
with a consumer credit field experiment. Econometrica, 77(6):1993–2008, 2009. URL
http://onlinelibrary.wiley.com/doi/10.3982/ECTA5781/abstract.
A. Khwaja, A. Mian, and A. Qamar. Bank credit and business networks. Working Paper,
2011. URL http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1763351.
A. I. Khwaja and A. Mian. Do lenders favor politically connected firms? rent provision in
an emerging financial market. The Quarterly Journal of Economics, 120(4):1371–1411,
2005a. URL http://qje.oxfordjournals.org/content/120/4/1371.short.
A. I. Khwaja and A. Mian. Unchecked intermediaries: Price manipulation in an emerging
stock market. Journal of Financial Economics, 78(1):203–241, 2005b. URL http:
//www.sciencedirect.com/science/article/pii/S0304405X0500053X.
A. I. Khwaja and A. Mian. Tracing the impact of bank liquidity shocks: Evidence from
an emerging market. The American Economic Review, pages 1413–1442, 2008. URL
http://www.jstor.org/stable/10.2307/29730128.
Ministry of Finance. Money and credit. Pakistan Economic Survey 2006, 2006.
S. Morris and H. S. Shin. Social value of public information. The American Economic Re-
view, 92(5):1521–1534, 2002. URL http://www.ingentaconnect.com/content/aea/
aer/2002/00000092/00000005/art00012.
D. Paravisini and A. Schoar. The information and agency effects of scores: Randomized
evidence from credit committees. Technical report, Working paper, November 9, 2012.
M. A. Petersen and R. G. Rajan. The benefits of lending relationships: Evidence
from small business data. The Journal of Finance, 49(1):3–37, 1994. URL http:
//onlinelibrary.wiley.com/doi/10.1111/j.1540-6261.1994.tb04418.x/full.
State Bank of Pakistan. Group liabilities in the cib report. Prudential Regulation, May
29th 2004.
J. C. Stein. Internal capital markets and the competition for corporate resources. The
Journal of Finance, 52(1):111–133, 1997. URL http://onlinelibrary.wiley.com/
doi/10.1111/j.1540-6261.1997.tb03810.x/abstract.
J. E. Stiglitz and A. Weiss. Credit rationing in markets with imperfect information.
The American economic review, 71(3):393–410, 1981. URL http://www.jstor.org/
stable/10.2307/1802787.
J. E. Stiglitz and A. Weiss. Incentive effects of terminations: Applications to the credit
and labor markets. The American Economic Review, 73(5):912–927, 1983. URL http:
//www.jstor.org/stable/10.2307/1814662.
J. Stroebel. The impact of asymmetric information about collateral values in mortgage
lending. Working paper, 2013.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 32
O. E. Williamson. Markets and hierarchies. New York, pages 26–30, 1975. URL
http://books.google.com/books?hl=en&lr=&id=pg-wGL12BjUC&oi=
fnd&pg=PA106&dq=williamson+markets&ots=vTVuhry98j&sig=
5cyDWz8eedTXmZZlkrKm52BjDAI.
O. E. Williamson. The economics of organization: The transaction cost approach. Amer-
ican Journal of Sociology, pages 548–577, 1981.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 33
8. Figures and Tables
Figure 4. Multiple overlapping groups
In the figure above there are three different firms and two directors but three differentgroups. Since a group is defined with respect to a firm (see definition 1), firm F1 onlyhas a direct link to firm F2. Therefore, F1’s group is only firms F1 and F2.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 34
Figure 5. The network of firm connections - using a cutoff of 25% equity
This is a network that shows interfirm relationships which only show connectionsbetween firms if they have at least one mutual director who owns more than 25% in bothfirms. Those firms with more interfirm relationships are colored in a darker red.The most visually striking aspect of this network is the lack of a single large component,and how the entire network is much less densely connected than the network compiledby merely mutual overlapping directors.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 35
Figure 6. The number of connections for each firm
The graph displays the number of connections each firm possess in December 2004 wherewe define two firms to be linked if they share a mutual director who owns 25% or moreof the firm’s equity.
Figure 7. The size of the components
This graph details the size of each component, where we define two firms to be linked ifthey share a mutual director who owns 25% or more of the firm’s equity. You can see thelargest component is only 14 firms.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 36
Figure 8. The histogram of log total firm borrowings
0.0
5.1
.15
.2.2
5D
ensi
ty
5 10 15 20Total firm borrowings
Total firm borrowings at baseline
This is the histogram of the log total firm borrowings in December 2004 for group firms. Since,
no loan below 500,000 Pk. Rupees was reported to the credit registry, there is a sharp cut-off at
6.2, which is approximately the log of 500.
Figure 9. The histogram of log loan size
0.0
5.1
.15
.2D
ensi
ty
6 8 10 12 14 16Funded loan size
Log loan sizes at baseline
This is the histogram of the log loan size in December 2004 for group firms. Since, no loan below
500 000 Pk. Rupees was reported to the credit registry, there is a sharp cut-off at 6.2, which is
approximately the log of 500.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 37
Figure 10. The histogram of log loan size for informed and uninformedbanking relationships
0.1
.2
5 10 15 20 5 10 15 20
Uninformed lending relationship Informed lending relationship
Den
sity
Funded loan sizeGraphs by whether the loan relationship is informative
Log loan sizes at baseline
This is the histogram of the log loan size in December 2004 separated by informed and
uninformed banking relationships. Since, no loan below 500,000 Pk. Rupees was reported to the
credit registry, there is a sharp cut-off at 6.2, which is approximately the log of 500. We provide
robustness tests to ensure the cutoff of small loans is not driving in our results.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 38
Figure 11. The histogram of total firm borrowings by lending relationship.
0.1
.2.3
5 10 15 20 5 10 15 20
Only one type of lending relationship Both types of lending relationship
Den
sity
Total firm borrowingsGraphs by whether the firm has both an informed and uniformed lending relationships
Total firm borrowings at baseline
This is the histogram of total firm borrowing in December 2004 separated by the firms we are
using for identification in section 5. We are only identifying the effects of the policy from the set
of firms who borrow from both an informed and uniformed lender, therefore, the right panel
demonstrates the distribution of total firm borrowings who have both an informed and
uninformed lending relationship. As you would expect the distribution of total borrowings for
firms who have both types of relationship are larger. This is partly mechanical effect as they
must have at least two loans. The larger total borrowings at baseline would suggest these firms
are also larger in general.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 39
Figure 12. The monthly coefficients for the difference in likelihood of aloan being renewed
This is the monthly coefficient for informed loans, when using ‘firm×date’, ‘firm×bank’ and
‘bank×date’ fixed effects. The light blue lines are point-wise 95% confidence intervals. It is
clearly evident there was a dramatic increase in the likelihood of a loan not being renewed for the
same firm when the borrowing relationship was an uninformed banking relationship.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 40
Figure 13. The monthly coefficients for the difference in log loan size
This is the monthly coefficient for informed loans, when using ‘firm×date’, ‘firm×bank’ and
‘bank×date’ fixed effects. The light blue lines are point-wise 95% confidence intervals. There
does not seem to be a pre-trend in the difference in log loan sizes but there is some evidence that
the loan sizes increased after the regulation change.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 41
Figure 14. The effect of the policy on the likelihood of the loan beingrenewed by different share thresholds.
This graph shows the effect of the policy according to different definitions of informativeness. In
particular, we create new dummies according to the interfirm relationship a bank observes.
Those relationships where the bank observes two firms with solely overlapping directors has
little to no effect on the likelihood a loan is renewed. Whereas if a lender observes two firms with
an interfirm relationship where there is a mutual director who owns more than 40% of the
company, those firms’ loans are 7.6% more likely to be renewed than if the lender only observed
one of those firms. This is the visual analogue of column 2 in table 8.
The 95% confidence interval is depicted with the straight lines, and the estimated coefficient is
the small blue box. All standard errors are clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 42
Figure 15. The effect of the policy on the likelihood of a loan beingrenewed by different number of firm’s who borrow from the same bank.
This graph shows as we increase the amount of firms who borrow from the same bank the effects
of the policy are greater. This is the visual analogue of column 2 in table 9, where we are plotting
the estimated coefficient and the standard errors from a regression of renewed loan on the
number of firm connections borrowing at bank b and the usual three second-order fixed effects.
The 95% confidence interval is depicted with the straight lines, and the estimated coefficient is
the small blue box.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 43
Figure 16. The effect of the policy on the likelihood of a loan beingrenewed by decile of firm size.
The graph shows the effects were greatest on the smallest firms (using December 2004 total
borrowings). This is the visual analogue of column 2 in table 10, where we are plotting the
estimated coefficient and the standard errors from a regression of renewed loan on decile (by firm
total borrowings in December 2004) interacted with informed loans and the usual three
second-order fixed effects.
The deciles are constructed using the total firm borrowings in December 2004, since those firms
who borrow small amounts don’t have multiple loans, the estimate for the bottom decile is not
identified.
The 95% confidence interval is depicted with the straight lines, and the estimated coefficient is
the small blue box.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 44
Figure 17. The effect of the policy on the likelihood of a loan beingrenewed by decile of group size.
The graph shows the effects were greatest on firms in the smallest groups (using December 2004
total borrowings). This is the visual analogue of column 1 in table 11, where we are plotting the
estimated coefficient and the standard errors from a regression of renewed loan on decile (by
group total borrowings in December 2004) interacted with informed loans and the usual three
second-order fixed effects. These results suggest that those firms in the very largest groups were
unaffected by the policy change. This is consistent with the suggestion that knowledge about the
largest groups is already well known in the banking sector.
The deciles are constructed using the total group borrowings in December 2004, since those
firms who borrow small amounts don’t have multiple loans, the estimate for the bottom decile is
not identified.
The 95% confidence interval is depicted with the straight lines, and the estimated coefficient is
the small blue box. Standard errors are clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 45
Figure 18. The monthly coefficients for the difference in total borrowingsbetween borrowers with informed lending relationships and those who donot.
This is the monthly coefficient for informed lender loans, when using ‘firm’, ‘group size specific
time trend’, ‘date FE×province FE’, ‘date FE×business sector FE’ controls. The light blue lines
are point-wise 95% confidence intervals. This graph shows that firms with an informed loan were
more more likely to receive relatively larger loans following the reform than group firms who did
not have a loan from an informed lender.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 46
Figure 19. The effect of the reform on a firm’s total borrowings by decileof firm borrowings in December 2004.
This graph shows the effects were greatest on the smallest firms (using December 2004 total
borrowings). This is the visual analogue of column 2 in table 15, where we are plotting the
estimated coefficient and the standard errors from a regression of change in total firm borrowings
on an informed lender interacted by each decile (by firm total borrowings in December 2004) and
various controls.
The 95% confidence interval is depicted with the straight lines, and the estimated coefficient is
the small blue box.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 47
Table 1. Summary statistics for the entire dataset in December 2004
Corporate Loans Consumers and Sole ProprietorsNumber of borrowers 11,395 86,053Percentage of total lending 60% 40%Overdue rates in December 2004 4.50% 7.11%
Mean Median S.D. Mean Median S.D.Loan size (in ’000 Pk. Rupees) 42,154 4,403 216,586 3,653 866 28,566Number of lending partners 1.73 1.00 1.79 1.06 1.00 0.29
In December 2004, the credit registry maintained details on the entire credit market if a loan was
greater than 500,000 Pk. Rupees. The table above demonstrates that largest lending was to
corporate firms, and there was a wide dispersion in loan sizes.
Table 2. Summary statistics for group firms in December 2004
Number of group firms 3,695Number of loans 7,250
Uninformed Loans Informed LoansNumber of loans 5,867 1,383
Mean MeanOverdue at baseline 2.77% 2.02%Loan size (in ’000s Pk. Rupees) 56,343 49,791Log loan size 9.28 9.43Public banks 7.0% 7.6%Private domestic commercial bank 56.5% 66.4%Non-Bank Fin. Corp. (NBFC) 28.2% 17.3%Foreign bank 8.3% 8.7%
A loan is defined at a firm-bank pair. Therefore, if a firm has multiple loans from the same
lender, it is classified as a single loan.
An informed loan is defined as a loan where at least two firms in the same group borrow from the
same lender. In the following regressions, the paper shows how a firm who has both a informed
and uninformed loan respond to the change in information.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 48
Table 3. Summary statistics for group firms who have both an informedand uninformed loan in December 2004
Number of group firms 449Number of loans 1,784
Uninformed Loans Informed LoansNumber of loans 1,035 749
Mean MeanOverdue at baseline 1.55% 1.33%Loan size (in ’000s Pk. Rupees) 45,667 65,162Log loan size 9.18 9.55Public banks 5.4% 5.0%Private domestic commercial bank 48.9% 63.0%Non-Bank Fin. Corp. (NBFC) 37.9% 23.1%Foreign bank 7.3% 8.9%
A loan is defined at a firm-bank pair. Therefore, if a firm has multiple loans from the same
lender, it is classified as a single loan.
Since the estimation results in section 5 are estimated from firms who have both an informed and
uninformed lending relationship, this is the set of firms which identify our main coefficient of
interest.
Table 4. The effect of the policy on the likelihood of a loan being renewedbetween informed and uninformed lending relationships
(1) (2) (3) (4) (5) (6)Informed 0.00213 0.0213 -0.00830 0.00453
(0.0150) (0.0135) (0.0135) (0.0126)Post*Informed 0.0562∗∗∗ 0.0734∗∗∗ 0.0722∗∗∗ 0.0767∗∗∗ 0.0517∗∗ 0.0543∗∗
(0.0121) (0.0244) (0.0241) (0.0242) (0.0217) (0.0218)Observations 269625 269625 269625 269625 269625 269625Firm&Date FE Yes N/A N/A N/A N/A N/ADate*Firm FE No Yes Yes Yes Yes YesBank FE Yes No Yes N/A N/A N/AFirm*Bank FE No No No Yes No YesBank*Date FE No No No No Yes Yes
Standard errors in parentheses∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
Each regression shows the likelihood of a informed loan to be renewed to be 5-7 pp more likely.
All the specifications 2-6 include a date×firm fixed effect, therefore, we identify the likelihood of
a informed loan to be renewed solely from the set of borrowers who have both a informed and
uninformed loan. All standard errors are clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 49
Table 5. The effect of the policy on the loan sizes between informed anduninformed lending relationships
Loan Size Loan Size Loan Size Loan Size Loan Size Loan SizeInformed 0.167∗∗ 0.449∗∗∗ 0.216∗∗∗ 0.227∗
(0.0761) (0.0944) (0.0728) (0.0709)Post*Informed 0.196∗∗∗ 0.111 0.120 0.125 0.0954 0.0916
(0.0480) (0.0869) (0.0854) (0.0830) (0.0786) (0.0775)Observations 269625 269625 269625 269625 269625 269625Firm&Date FE Yes N/A N/A N/A N/A N/ADate*Firm FE No Yes Yes Yes Yes YesBank FE Yes No Yes N/A N/A N/AFirm*Bank FE No No No Yes No YesBank*Date FE No No No No Yes Yes
Standard errors in parentheses∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
Each regression shows the likelihood of a informed loan to be to be 9-13% to be larger following
the reform. All the specifications 2-6 include a date×firm fixed effect, therefore, we identify the
difference in loan size solely from the set of borrowers who have both a informed and uninformed
loan. All standard errors are clustered at the level of the component.
Table 6. The effect of the policy on the loan sizes between informed anduninformed lending relationships restricting to only those loans that lasteduntil June 2008 or December 2008.
Loan Size Loan SizePost*Informed 0.120 0.0698
(0.107) (0.120)Observations 74131 74417Date cutoff June 2008 Dec 2008Date*Firm FE Yes YesFirm*Bank FE Yes YesBank*Date FE Yes Yes
Standard errors in parentheses∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
This set of regressions restricts attention to only those firm-banks pairs which last until June
2008 or December 2008. The results are very similar to those from table 5, suggesting the effect
of the policy is not being driven by firms dropping out of the sample. All standard errors are
clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 50
Table 7. The effect of the policy on the loan sizes between informed anduninformed lending relationships - censoring at baseline.
Loan Renewed Loan Renewed Loan Renewed Loan RenewedPost*Informed 0.0543∗∗ 0.0495∗∗ 0.0356 0.0478∗
(0.0218) (0.0228) (0.0248) (0.0267)Observations 269625 258814 213599 181867Cutoff $8500 $12,750 $17,000 $21,250Date*Firm FE Yes Yes Yes YesFirm*Bank FE Yes Yes Yes YesBank*Date FE Yes Yes Yes Yes
Standard errors in parentheses∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
The SBP had a policy of only obliging banks to report a loan’s details if the firm’s totalloan outstanding was greater than 500,000 Rupees ($8,500). This could lead to a bias inour results since loans which were initially just above the cutoff may remain active butpartially repaid to be subsequently below 500,000 Rupees and incorrectly categorized asa non-renewed loan. This table omits firms which had a loan amount close to the cutoffin December 2004. Column 1 is the baseline results (Table 4), column 2 omits firms whohad a loan outstanding below $12,750 in December 2004 (50% larger than the SBPcutoff), column 3 omits firms who had a loan outstanding below $17,000 in December2004 (100% larger than the cutoff), and column 4 omits firms who had a loanoutstanding below $21,250 in December 2004 (150% larger than the cutoff). All theestimates are similar to those in column 1 suggesting the cutoff is not leading to anoticeable bias in the results. All standard errors are clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 51
Table 8. The effect of greater shareholding on the effect of the policy onloan sizes and likelihood of renewing the loan.
Loan Size Loan RenewedPost*Informed 0% 0.0397 0.0248
(0.0716) (0.0202)Post*Informed 10% 0.0980 0.0440∗
(0.0936) (0.0234)Post*Informed 25% 0.170∗ 0.0555∗∗
(0.0993) (0.0275)Post*Informed 40% 0.0218 0.0767∗∗
(0.122) (0.0328)Observations 269625 269625Date*Firm FE Yes YesFirm*Bank FE Yes YesBank*Date FE Yes Yes
Standard errors in parentheses∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
Construction of the ‘informed 40%’ variable is the same as the procedure in the paper forconstructing the benchmark bank-firm pairs, except using a 40% cutoff for directorialshareholding. We construct the ‘informed 25%’ variable to be those firm-bank pairs who
would be defined as informed using the usual definition with a 25% director shareholder cutoff -
except omitting those firm-bank pairs that were already labeled as ‘informed 40%’. A similar
procedure is used to construct the 10% and 0% informed variables. All standard errors are
clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 52
Table 9. The effect of the greater number of firms borrowing from thesame institution on loan sizes and likelihood of renewing the loan.
Loan Size Loan RenewedPost*2 Informed firms 0.0554 0.0431∗
(0.0804) (0.0228)Post*3 Informed firms 0.212 0.111∗∗∗
(0.143) (0.0356)Post*3+ Informed firms 0.446∗ 0.0913∗
(0.229) (0.0511)Observations 269625 269625Date*Firm FE Yes YesFirm*Bank FE Yes YesBank*Date FE Yes Yes
Standard errors in parentheses∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
Construction of the ‘# informed firms’ variable is done through adding the total number of firms
in a firm’s group who borrow from the same lender. Therefore, these regressions suggest a bank
was more willing to renew a loan, the greater the number of group firms the lender lent to in
December 2004. All standard errors are clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 53
Table 10. The effect of the policy by decile of firm size
Loan Size Loan RenewedPost*Informed*2nd Decile by Firm Size 0.461∗∗ 0.543∗∗∗
(0.232) (0.127)Post*Informed*3rd Decile by Firm Size 0.247 0.166
(0.242) (0.108)Post*Informed*4th Decile by Firm Size -0.0207 -0.122
(0.207) (0.122)Post*Informed*5th Decile by Firm Size 0.112 0.0607
(0.212) (0.0783)Post*Informed*6th Decile by Firm Size 0.356 0.109
(0.211) (0.0738)Post*Informed*7th Decile by Firm Size 0.0125 0.0262
(0.215) (0.0686)Post*Informed*8th Decile by Firm Size 0.134 0.158∗∗∗
(0.146) (0.0519)Post*Informed*9th Decile by Firm Size 0.102 0.0225
(0.174) (0.0446)Post*Informed*10th Decile by Firm Size -0.00955 -0.00843
(0.152) (0.0355)Observations 269625 269625Date*Firm FE Yes YesFirm*Bank FE Yes YesBank*Date FE Yes Yes
Standard errors in parentheses∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
Each decile is constructed by computing the total firm borrowings in December 2004. The effect
on the smallest decile is not identified in the data since the smallest borrowers in December 2004
only had one loan and the identification relies on a borrower having at least two loans. All
standard errors are clustered at the level of the component. The results for column 2 are shown
in figure 16.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 54
Table 11. The effect of the policy by decile of group size
Loan Renewed Loan SizePost*Informed*2nd Decile by Group Size 0.191 -0.112
(0.242) (0.397)Post*Informed*3rd Decile by Group Size 0.0394 -0.158
(0.139) (0.227)Post*Informed*4th Decile by Group Size -0.0168 -0.106
(0.109) (0.252)Post*Informed*5th Decile by Group Size 0.187∗∗ 0.460∗∗∗
(0.0864) (0.164)Post*Informed*6th Decile by Group Size -0.00495 0.0781
(0.0690) (0.214)Post*Informed*7th Decile by Group Size 0.158∗∗∗ 0.0999
(0.0589) (0.195)Post*Informed*8th Decile by Group Size 0.123∗∗ 0.297∗
(0.0605) (0.177)Post*Informed*9th Decile by Group Size 0.0249 -0.0769
(0.0438) (0.159)Post*Informed*10th Decile by Group Size 0.0155 0.0967
(0.0368) (0.152)Observations 269625 269625Date*Firm FE Yes YesFirm*Bank FE Yes YesBank*Date FE Yes Yes
Standard errors in parentheses∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
Each decile is constructed by computing the total group borrowings in December 2004. The
effect on the smallest decile is not identified in the data since the smallest borrowers in December
2004 only had one loan and the identification relies on a borrower having at least two loans. The
results show that the effect of the reform was largest on those firms who belonged to the smallest
group. This is consistent with the view that there is common knowledge about a firm’s group for
the very largest groups.
All standard errors are clustered at the level of the component. The results for column 2 are
shown in figure 17.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 55
Table 12. Firms who were overdue at baseline were more likely to renewtheir informed loan.
Loan Size Loan RenewedPost*Informed 0.0681 0.0477∗∗
(0.0778) (0.0222)Post*Informed*Firm Overdue 0.472∗ 0.133∗
(0.278) (0.0691)Observations 269625 269625Date*Firm FE Yes YesFirm*Bank FE Yes YesBank*Date FE Yes Yes
Standard errors in parentheses∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
This is suggestive evidence that those firms who had been overdue in the past were more likely to
renew their loans at an informed banking relationships. This suggests that those firms for may be
classified as the most risky were also the ones most likely to have their loans renewed by an
informed lender. All standard errors are clustered at the level of the component.
Table 13. The difference in the likelihood of a loan going overdue
Overdue OverdueInformed 0.0163∗∗∗
(0.00518)Post*Informed -0.0287∗∗∗ -0.00719
(0.00666) (0.0141)Observations 209962 209962Firm&Date FE Yes N/ADate*Firm FE No YesBank FE Yes N/AFirm*Bank FE No YesBank*Date FE No Yes
Standard errors in parentheses∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
Caution should be taken when taking inference from these regressions. As described in the text,
this is not estimating the causal impact of the regulation change on the likelihood of a loan
becoming overdue, since in table 4, we showed uninformed loans were more likely to be not
renewed. Column 1 does not include firm interacted with date fixed effects. Therefore, column 1
suggests the loans that were renewed by uninformed lenders were more likely to default. Column
2 which includes the firm interacted with date fixed effect suggests that those firms for whom the
informed and uninformed lender renewed, there was no differential in default.
All standard errors are clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 56
Table 14. Difference in total loan sizes between firms who have an in-formed lender and those who do not.
Total Loans Total Loans Total Loans Total LoansPost*Informed lender 0.130∗∗∗ 0.134∗∗∗ 0.113∗∗ 0.123∗∗
(0.0447) (0.0458) (0.0464) (0.0476)Constant 9.845∗∗∗
(0.0470)Observations 161744 155496 149248 145684Firm FE Yes Yes Yes YesDate FE Yes N/A N/A N/AGroup Size Specific Time Trend Yes Yes Yes YesBusiness Sector FE*Date FE No Yes No YesProvince FE*Date FE No No Yes Yes
Standard errors in parentheses∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
The difference-in-difference estimates suggest that those firms with an informed lenderwere able to borrow 11%-14% more following the reform than those firms with noinformed lender. Column 1 shows the estimates with firm fixed effects and a group sizespecific time trend. Column 2 shows the estimates when including business sectorinteracted with date fixed effects . Column 3 shows the estimates when includingprovince interacted with date fixed effects. Lastly column 4 shows the estimates wherewe include all interacted fixed effects. We are missing the data on the province andbusiness sector for a small fraction of the firms, therefore, they are omitted from thoseregressions which include province or business sector fixed effects. All standard errors are
clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 57
Table 15. The difference in total loan borrowings by firm decile andwhether they had an informed lending relationship.
Total Loans Total LoansPost*Informed*1st Decile 0.764∗∗∗ 0.756∗∗∗
(0.118) (0.137)Post*Informed*2nd Decile 0.253∗∗ 0.290∗∗
(0.103) (0.121)Post*Informed*3rd Decile 0.310∗∗∗ 0.318∗∗∗
(0.0957) (0.101)Post*Informed*4th Decile -0.274∗∗∗ -0.311∗∗∗
(0.104) (0.119)Post*Informed*5th Decile 0.217∗∗ 0.173
(0.102) (0.111)Post*Informed*6th Decile 0.0678 0.0708
(0.0988) (0.101)Post*Informed*7th Decile -0.0511 -0.0282
(0.129) (0.128)Post*Informed*8th Decile -0.0714 -0.100
(0.134) (0.136)Post*Informed*9th Decile 0.267∗∗ 0.257∗∗
(0.105) (0.105)Post*Informed*10th Decile 0.0340 0.0728
(0.132) (0.143)Constant 9.845∗∗∗
(0.0458)Observations 161744 145684Firm FE Yes YesDate FE Yes N/AGroup Size Specific Time Trend Yes YesBusiness Sector*Date FE No YesProvince*Date FE No Yes
Standard errors in parentheses∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
These regressions interact a dummy variable for each decile (by total firm borrowings in
December 2004) with the post×informed lender variable. Hence, we are estimating the effect of
the policy separately for each decile. The estimates suggest that the largest effects were on the
smallest firms, which is similar to the results shown in section 5.2.3. Figure 19 plots the results
in column 2. All standard errors are clustered at the level of the component.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 58
9. Appendix
9.1. Proof of Proposition 1. A Perfect Bayesian Equilibrium of the game is:
(informed lender) kLI = 0, kHI = kH∗I > 0
(uninformed lender) kN = k∗N ≥ 0
(firm) sif = (Acc, Acc)
Proof. To prove the exist a PBE we will first solve for the firm’s weakly dominant strategy,
and then we shall proceed to solve for the PBE under different parameter values.
To solve for the firm’s weakly dominant strategy recall the firm’s utility function:
Ui(k,R) =Xi
1 + k(A−R)k
The firm’s utility function is strictly increasing in capital, k, for all non-negative k since we
assume the productivity parameter A is strictly greater than the interest rateR. Therefore,
the firm’s weakly dominant strategy is to accept all loan offers from both lenders.
Now let us consider the informed lender’s optimal strategy if the uninformed lender is
offering kN and the firm accepts all loan offers. The informed lender wishes to maximise:
πiI(kiI , kN , Xi) = max
[Xi
1 + kiI + kNR− ρ
]kiI
Since we assumed that RXLρ < 1, we know that the informed lender’s strictly dominant
strategy having observed a low borrower type will be kLI = 0 since expected profits will
always be strictly negative if she ever offers a non-zero loan size.
Now let us consider the case where the informed lender observes a high borrower type and
let us assume that the following condition holds:
(2) γ
(RXL
ρ
)+ (1− γ)
(RXH
ρ
)0.5
< 1
Then there exists a PBE:
(informed lender) kL∗
I = 0, kH∗
I = kmaxH =
(RXH
ρ
)0.5
− 1 > 0
(uninformed lender) k∗N = 0
(firm) sif = (Acc, Acc)
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 59
To verify this is a PBE, the informed lender’s marginal profit condition is:
(3)∂π
∂kI= RXH
(1 + kN )
(1 + kN + kHI )2− ρ = 0
And since the profit function is concave in kiI the local maximum is sufficient for the
lenders’ optimal strategy given the other lender’s strategy. If the uninformed lender
chooses kN = 0, then the solution to equation 3 is kHI = kmaxH . To check this is a
PBE, we need to check the uninformed lender’s optimal strategy.
The marginal profit condition for the uninformed lender is:
∂π
∂kN= γRXL
1
(1 + kN )2+ (1− γ)RXH
(1 + kHI )
(1 + kN + kHI )2− ρ
If we substitute in the conjectured optimal informed lender’s strategy, this becomes:
(4)∂π
∂kN
∣∣∣∣kH=kmax
H
= γRXL1
(1 + kN )2+ (1− γ)RXH
(RXHρ
)0.5(kN +
(RXHρ
)0.5)2 − ρ
Similar to the informed lender’s problem, the uninformed lender’s profit is concave in kN .
Therefore, if we can show the uninformed lender’s marginal profit is negative at kN = 0
we know the uninformed lender’s optimal strategy given the informed lender’s strategy is
to offer kN = 0.
Plugging in kN = 0 into equation 4 gives:
∂π
∂kN
∣∣∣∣kN=0,kH=kmax
H
= γRXL + (1− γ)RX0.5H − ρ
Using the assumption in equation 2, we know this is strictly less than zero. Therefore,
the uninformed lender’s optimal strategy given the informed lender’s strategy is to offer
kN = 0.
To complete the proof for all feasible values of XH and XL we will solve the uninformed
and informed lender’s maximization problems assuming the constraints that each lender
does not have a negative loan size or a negative expected profits do not bind. We will
then show for those parameter values where the constraints do bind the equilibrium will
be such that k∗N = 0, kL∗
I = 0, kH∗
I = kmaxI .
Recall the informed lender’s marginal profit condition:
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 60
(5)∂π
∂kI= RXH
(1 + kN )
(1 + kN + kHI )2− ρ
Recall the uninformed lender’s marginal profit condition:
(6)∂π
∂kN= γRXL
1
(1 + kN )2+ (1− γ)RXH
(1 + kHI )
(1 + kN + kHI )2− ρ
Assuming no constraints bind, we can set equations 5 and 6 equal to zero and solve for
kHI and kN as a function of the parameters (the first order conditions are sufficient since
both lenders’ profits are concave in their respective capital offers). First we show that the
solution of these two equations imply that kH∗
I > 0.
To reduce notation let us define Y ≡ 1 + kN , ZH ≡ RXHρ and ZL ≡ RXL
ρ .
Then we can rewrite the two marginal profit conditions (FOCs) as:
ZHY = (kI + Y )2(7)
γZL1
Y 2+ (1− γ)ZH
(1 + kHI )
(kHI + Y )2= 1(8)
Plugging in (kI + Y )2 from equation 7 into equation 8. We have the equation:
(9) γZL1
Y 2+ (1− γ)ZH
(1 + kHI )
ZHY= 1
Simplifying and rearranging equation 9:
(10) Y 2 − (1− γ)(1 + kHI )Y − γZL = 0
Solving for Y :
(11) Y =(1− γ)(1 + kHI )±
√(1− γ)2(1 + kHI )2 + 4γZL
2
We are interested in kN > 0 (the maximum), so we can restrict ourselves to the positive
root of Y .
Rearranging equation 7:
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 61
(12) Y 2 + (2kHI − ZH)Y + (kHI )2 = 0
Subtracting 10 from 12 gives:
(13)(2kHI − ZH + (1− γ)(1 + kI)
)Y + (kHI )2 + γZL = 0
We can substitute Y from equation 11 into 13. Equation 13 is continuous in kHI , the left
hand size of 13 is less than zero if kHI = 0, and the left hand side of13 tends to infinity
as kHI goes to infinity, therefore there must exist a positive kHI such that both FOCs are
satisfied. Since, there exists a kHI which satisfy the two FOCs, then there must also exist
a corresponding kN .
However, the choice of kN , which satisfies the FOC may not be feasible. In particular, at
those parameter values kN may be less than zero. We now show that kN is always greater
than zero from the FOC conditions.
Recall equation 11:
Y =(1− γ)(1 + kHI )±
√(1− γ)2(1 + kHI )2 + 4γZL
2
We want to show that Y ≡ 1+kN is always greater than one. Since, Y is increasing in kHI ,
and we have shown kHI is always greater than zero, then if we show this equation holds
when kHI = 0, it must hold for all parameter values. Therefore:
Y =(1− γ)(1 + kHI ) +
√(1− γ)2(1 + kHI )2 + 4γZL
2> 1
Substituting in kHI = 0 and rearranging, we obtain:
√(1− γ)2 + 4γZL > (1 + γ)
Square both sides (both sides are positive) and rearrange:
ZL >1
2
From our assumptions on the parameter space, this is always true. Therefore both the kHIand kN that satisfy the two FOCs are greater than zero.
Finally to complete the proof, we must show what happens when the uninformed lender’s
profit condition is not satisfied. We need the following preliminaries: (i) the informed
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 62
lender’s best response function is decreasing in kN for all feasible values of kN , (ii) the
uninformed lender’s best response function is decreasing in kHI and (iii) in equilibrium
kHI >kN .
Proof of preliminary (i)
Totally differentiating equation 5 with respect to kN and rearranging:
(14)dkHIdkN
=Z0.5H
2(1 + kN )0.5− 1 < 0
Therefore, the optimal kHI is decreasing for all non-negative values of kN since we assumed
ZH < 4.
Proof of preliminary (ii)
Totally differentiating the uninformed lender’s FOC condition (equation 8) with respect
to kHI gives:
−2γZL1
Y 3
dY
dkHI+
(1− γ)ZH
(kHI + Y )4
[(kHI + Y )2 − 2
(1 +
dY
dkHI
)(kHI + Y )(1 + kHI )
]= 0
Rearranging and simplifying:
dY
dkHI= (kHI + Y )2 − 2(kHI + Y )(1 + kHI )
dkN
dkHI= (kHI + Y )(kN − 1− kHI )
Therefore, if kHI is greater than kN then the uninformed lender’s choice of kN is decreasing
in kHI .
Proof of preliminary (iii)
Follows from the proof of proposition 2.
Therefore, now we have shown that kN is decreasing in kHI , and kHI is decreasing in kN .
We know if the FOC suggest a solution with negative expected profits for the uninformed
lender, the uninformed lender will always set kN = 0. This follows from (i) the informed
lender’s optimal response will be to increase her choice of kHI if the uninformed lender’s
choice of kN decreases, and (ii) if the informed lender’s choice of kHI increases, the un-
informed lender will always wish to decrease her choice of kN – ultimately, this leads to
kN = 0.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 63
Finally, we need to check the informed lenders expected profits are always greater than
zero. There are two cases: (i) if the uninformed lender offers a positive loan size and (ii)
the uninformed lender does not enter.
Case (i):
If the uninformed lender has a positive loan size then the following condition must hold:
(15) RXH1
(1 + k∗N + kH∗
I )2− ρ ≥ 0
Recalling the informed lender’s profit function:
E[πHI (kH
∗I , kN )
]= RXH
kH∗
I
(1 + k∗N + kH∗
I )2− ρkH∗I
If equation 15 is satisfied then the informed lender’s profits must be greater than zero.
Case (ii):
The informed lender profits must similarly be greater than zero since RXHρ is assumed to
be greater than 1, therefore, the informed lender can always make a positive profit.
To sum, we have shown there exists a solution to first order conditions when we do not
restrict profits to be non-negative or loan size to be positive. Then we have shown a PBE
there exists a PBE for those parameter values such that the first order conditions give a
non-feasible solution.
Further, it can be shown there is no mixed strategy PBE in this model. �
9.2. Proof of Proposition 2. The informed lender will lend more in expectation than
the uninformed lender:
∆ ≡ (1− γ)kH∗
I − k∗N > 0
Proof. There are two possible cases, (i) k∗N = 0 and (ii) k∗N > 0.
Case (i): k∗N = 0
Consider the informed lender’s FOC:
(16) RXH(1 + kN )
(1 + kN + kHI )2− ρ = 0
Using k∗N = 0 and the assumption that RXHρ > 1 (that it is profitable to offer loans to the
high type), then the equilibrium kH∗
I > 0 and π∗I > 0. It follows ∆ > 0.
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 64
Case (ii): k∗N > 0
In this situation both the lenders’ FOCs must hold. We will first solve for the equilibrium
k∗I and k∗N which satisfy the two FOCs, and then using a proof by contradiction we will
show that (1− γ)k∗I − k∗N > 0.
To reduce notation let us define Y ≡ 1 + kN , ZH ≡ RXHρ and ZL ≡ RXL
ρ .
Recall equation 11
(17) Y =(1− γ)(1 + kI)±
√(1− γ)2(1 + kI)2 + 4γZL
2
We are interested in kN > 0 (the maximum), so we can restrict ourselves to the positive
root of Y .
We want to show that ∆ ≡ (1− γ)kI − kN ≡ 1 + (1− γ)kI − Y > 0. We will complete the
proof by contradiction. Assume ∆ ≤ 0 then:
1 + (1− γ)kI ≤ Y
1 + (1− γ)kI ≤(1− γ)(1 + kI)±
√(1− γ)2(1 + kI)2 + 4γZL
2(18)
In equation 18 we have substituted Y from equation 17.
Rearranging and simplifying equation 18:
(1− γ)kI + (1 + γ) ≤√
(1− γ)2(1 + kI)2 + 4γZL
If we square both sides:
((1− γ)kI + (1 + γ))2 ≤ (1− γ)2(1 + kI)2 + 4γZL
Simplifying:
(19) 4γ ≤ 4γZL − 4γkI
Since by assumption 1 > ZL and kI ≥ 0 then 19 is false. Thereby completing the proof. �
9.3. Proof of Proposition 3. If the uninformed continues to lend, the uninformed lender
will have greater rates of default than the informed lender:
∆D ≡ E[DN −DI |kN > 0
]> 0
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 65
Proof. We restrict attention to this equilibrium where the uninformed lend continues to
lend (since the uninformed lend cannot have a defaults if she does not lend).
In equilibrium, the informed lender’s default rate is:
(20) E(DI) =
(1− XH
1 + kH∗
I + k∗N
)The uninformed lender’s default rate is:
(21) E(DN |k∗N > 0) = γ
(1− XL
1 + k∗N
)+ (1− γ)
(1− XH
1 + kHI + k∗N
)Therefore to complete the proof we need to show that, equation 21 is greater than equation
20:
γ
(1− XL
1 + k∗N
)+ (1− γ)
(1− XH
1 + kH∗
I + kN
)>
(1− XH
1 + kH∗
I + k∗N
)XL
1 + k∗N<
XH
1 + kH∗
I + k∗N(22)
Recalling that the informed lender must make non-negative profits, then the RHS of
inequality 22 must be greater than or equal to Rρ . From our initial assumptions XL must
be less than equal Rρ . k∗N must be greater than zero. Therefore inequality 22 holds. �
9.4. Proof of Proposition 4. As we reduce XL, the expected difference between how
much the informed and uniformed lender offer is increasing.
∆ ≡ (1− γ)kI − kN is decreasing in XL
Proof. A change in XLonly affects the informed lender through it’s effect on the un-
informed lender’s choice of kN (formally, partially differentiating the informed lender’s
profit maximization condition with respect to XLis zero).
Partially differentiating the uninformed lender’s FOC condition with respect to XL:
γR1
(1 + kN )2− 2γRXL
1
(1 + kN )3∂kN∂XL
− 2(1− γ)RXH(1 + kHI )
(kHI + Y )3∂kN∂XL
= 0
Therefore it follows:∂kN∂XL
≥ 0
Using the proof of proposition 1, we know that the informed lender’s choice of kI is decreas-
ing in kN , similarly the uninformed lender’s choice of kN is decreasing in kI . Consequently,
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 66
if XL rises, holding kHI constant, we know that kN rises, but then this itself causes kHI to
fall, which in turn causes kN to rise more and so on. �
9.5. Proof of Corollary 1. A Perfect Bayesian equilibrium of the game is:
(Informed lender) k̃LI = 0, k̃HI = k̃H∗I > 0
(Uninformed lender) kN = k∗N ≥ 0
(Firm) sif = (Acc, Acc)
Proof. To be completed
The proof is very similar to proposition 1. �
9.6. Proof of Proposition 5. As the informed lender’s signal of the borrower improves,
the difference in expected lending increases. Formally:
∆̃ ≡ [(1− q)γ + q(1− γ)] k̃I − kN is increasing in q
Proof. To be completed
The proof is very similar to proposition 4. �
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 67
Figure 20. Credit Worthiness Report from 2004
A copy of the information provided by the SBP about a prospective borrower in 2004
PUBLIC INFORMATION AND ASYMMETRICALLY INFORMED LENDERS 68
Figure 21. Credit Worthiness Report from 2012
STATE BANK OF PAKISTANCONSUMER PROTECTION DEPARTMENT
Corporate Credit Information Report
Code:
Name:
Report Ref. Report Date: 06/12/2012
Corporate Profile
Code:
Present Address: ,
Previous Address:
Consolidated Credit Exposure As on
Outstanding Liabilities
Fund Based 34,316,277
Non Fund Based 0
Overdues
Past Dues 90 Days
Past Dues 365 Days
0
0
Name:
Amount Under Litigation
Date and amount of Recovery
No of times Rescheduling/
0
0
0
31/10/2012
Group Entities of the Borrower
Restructuring During last five years1
Writes-offs (During last ten years)
Code # Name Of Entity
Credit Enquiries
Enquiring Financial Institute Enquiry Date
PRIVATE SECTOR COMMERCIAL BANK 22/11/12
PRIVATE SECTOR COMMERCIAL BANK 28/08/12
PRIVATE SECTOR COMMERCIAL BANK 31/05/12
PRIVATE SECTOR COMMERCIAL BANK 28/02/12
PRIVATE SECTOR COMMERCIAL BANK 02/12/11
Remarks
The Information contained in this report has been compiled from the data provided by the finacial institutions and does not represent the opinion ofState Bank of Pakistan as to credit worthiness of the subject. Hence State Bank of Pakistan cannot assure any liability to the accuracy orcompleteness of the information. The information contained in this report is supplied on a confidential basis to you and shall not be disclosed to anyother person
Disclaimer:
A copy of the information provided by the SBP about a prospective borrower in 2012.
The Presidents/Chief Executives May 29, 2004
All Banks/ DFIs/NBFCs
Dear Sirs/Madam,
GROUP LIABILITIES IN THE CIB REPORTS
The definition of “Group” for the purpose of CIB report as notified vide Circular No. SBP/CIB-23/94 dated August 22, 1994 has been reviewed. It has been decided that in the CIB report, the grouping of borrowers shall now be done on the basis of following criteria:
a) The names of the “Group” companies shall be reported by the financial institutions according to the definition of “Group” as contained in Definition No. 14 of the Prudential Regulations (PRs) for Corporate /Commercial Banking. Thus, the onus for correct formation of the group as per definition given in the Prudential Regulations will be on Banks/DFIs/NBFCs.
b) Banks/DFIs/NBFCs will ensure that while determining the group relationships in terms of criteria prescribed in the PRs, they should not consider the foreign national directors, directors of companies under liquidation, and nominee directors of the followingentities/agencies:
Foreign Controlled Entities. · Banks/DFIs. · Public Sector Enterprises. · Federal/Provincial Government. · Private Sector Enterprises’ nominee directors on the Board of Public Sector Enterprises.
c) The definition of the Group as contained in PRs shall, however, be not applicable in the context of Government owned / controlled entities notwithstanding the fact these are listed or unlisted.
2) Since reflection of negative information in the credit report of any party adversely impacts its relationships with its lending institutions, therefore, Banks/DFIs/NBFCs are advised to be very careful while reporting the names of group entities in the CIB data. In case any party disputes the group relationship, the reporting Bank/DFI/NBFC should be able to defend its position with documentary evidence.
3) The above changes in the grouping criteria of companies have necessitated collection of certain additional data from the financial institutions. Therefore, existing formats of CIB data collection (viz. CIB-I, II and III) circulated vide BSD Circular No.4 dated 25th February, 2003 have been revised. The revised formats for CIB I, II and III are enclosed
as Annexure-I. Banks/DFIs/NBFCs are advised to start collecting the additional information called for in the above formats from their clients and form the groups as per definition of the “Group” given in Prudential Regulations. The SBP will start collecting the monthly CIB data on the revised formats after revising its data capturing software application, for which banks/DFIs/NBFCs will be advised in due course.
Please acknowledge receipt.
Yours faitfully
(Jameel Ahmad)
Director